Geosci. Model Dev., 4, 919–941, 2011 www.geosci-model-dev.net/4/919/2011/ Geoscientific doi:10.5194/gmd-4-919-2011 Model Development © Author(s) 2011. CC Attribution 3.0 License.

The Met Office Unified Model Global Atmosphere 3.0/3.1 and JULES Global Land 3.0/3.1 configurations

D. N. Walters, M. J. Best, A. C. Bushell, D. Copsey, J. M. Edwards, P. D. Falloon, C. M. Harris, A. P. Lock, J. C. Manners, C. J. Morcrette, M. J. Roberts, R. A. Stratton, S. Webster, J. M. Wilkinson, M. R. Willett, I. A. Boutle, P. D. Earnshaw, P. G. Hill, C. MacLachlan, G. M. Martin, W. Moufouma-Okia, M. D. Palmer, J. C. Petch, G. G. Rooney, A. A. Scaife, and K. D. Williams Met Office, FitzRoy Road, Exeter, EX1 3PB, UK Received: 20 April 2011 – Published in Geosci. Model Dev. Discuss.: 17 June 2011 Revised: 26 September 2011 – Accepted: 14 October 2011 – Published: 26 October 2011

Abstract. We describe Global Atmosphere 3.0 (GA3.0): configurations (Cullen, 1993). For nearly twenty years, this a configuration of the Met Office Unified Model (MetUM) has provided the Met Office and its collaborators with the developed for use across climate research and weather pre- benefit of a single technical infrastructure, dynamical core diction activities. GA3.0 has been formulated by con- and collection of atmospheric parametrizations and has al- verging the development paths of the Met Office’s weather lowed them to develop modelling systems efficiently with- and climate global atmospheric model components such out the duplication of effort on common programming tasks. that wherever possible, atmospheric processes are modelled This technical efficiency also extends to the scientific devel- or parametrized seamlessly across spatial resolutions and opment of the model. Climate models are being run at ever timescales. This unified development process will provide higher resolution and applied to problems beyond the predic- the Met Office and its collaborators with regular releases of tion of mean climate states, such as the frequency and sever- a configuration that has been evaluated, and can hence be ap- ity of extreme weather events (Huntingford et al., 2003). This plied, over a variety of modelling regimes.´ We also describe requires the accurate modelling of mesoscale systems, which Global Land 3.0 (GL3.0): a configuration of the JULES com- has long been the remit of NWP forecasting. Conversely, to munity land surface model developed for use with GA3.0. consistently improve the quality of weather forecasts, global This paper provides a comprehensive technical and scien- and regional NWP models are starting to employ more com- tific description of the GA3.0 and GL3.0 (and related GA3.1 plexity and modelling elements of the Earth System that have and GL3.1) configurations and presents the results of some previously been the reserve of climate modelling, such as initial evaluations of their performance in various applica- atmospheric composition (Milton et al., 2008) and detailed tions. It is to be the first in a series of papers describing modelling of the water cycle (Balsamo et al., 2011). Collab- each subsequent Global Atmosphere release; this will pro- oration between climate and weather scientists and the use vide a single source of reference for established users and of a common atmospheric model aids users of the MetUM to developers as well as researchers requiring access to a cur- make these developments in an efficient manner. rent, but trusted, global MetUM setup. As well as the efficiency of central code development, us- ing the MetUM allows the Met Office to take advantage of the recognised synergies between climate and NWP mod- 1 Introduction elling (Martin et al., 2010; Senior et al., 2010). By study- The Met Office Unified Model™ (MetUM) is a highly flex- ing the same model formulation across a range of tempo- ible atmospheric model that was designed from its inception ral scales one can distinguish systematic biases that develop to be used for both climate research and Numerical Weather on very short timescales due to errors in rapidly responding Prediction (NWP) activities in both global and limited area parametrizations (e.g. clouds and precipitation) from those that develop due to more slowly evolving elements such as the soil moisture content in the land surface model. Study- Correspondence to: D. N. Walters ing the same model across different resolutions also allows (david.walters@metoffice.gov.uk) one to investigate the sensitivity of processes such as tropical

Published by Copernicus Publications on behalf of the European Geosciences Union. 920 D. N. Walters et al.: MetUM GA3.0/3.1 and JULES GL3.0/3.1 configurations cyclone formation to the resolved spatial scales (Heming, longstanding differences from the climate model; we label 2010) and the efficiency of parametrization schemes in mod- these GA3.1 and GL3.1. These differences and a discussion elling unresolved processes such as atmospheric convec- of the motivations behind them are included in Sect. 5, Fi- tion (Lean et al., 2008). Finally, verifying a model against nally, we present some initial results from assessments of the observations, atmospheric analyses and climatologies greater configurations in Sect. 6 and summarise our current position exposes its deficiencies and leads to a more robust formula- in Sect. 7. tion. For this reason, the Met Office has kept the develop- ment paths of the global forecast models within its opera- 2 The “Global Atmosphere” configuration tional NWP suite and the Met Office Hadley Centre’s cli- 2.1 Definition of the Global Atmosphere mate and seasonal forecasting models roughly aligned. Ma- jor developments introduced in one set of models are usu- The principle of the Global Atmosphere configuration fol- ally adopted by the other after only a short delay (e.g. Martin lows that set out for the atmospheric component of the et al., 2006; Milton et al., 2005). This process is hampered, HadGEM3 family in Hewitt et al. (2011), but extended however, when developments that become well established to include short-range NWP configurations. The Global in one context prove deficient in another, which can lead to Atmosphere is a single choice of dynamical core, atmo- either the inefficient development of a more general solu- spheric parametrizations, and options therein, applied to tion or a divergence of the configurations. To improve this an atmospheric MetUM component which may well itself situation, the Met Office has recently formalised the coor- be part of a larger system. For example, the global cli- dination of these development paths by merging them into mate modelling system HadGEM3-AO currently uses a Me- a single “Global Atmosphere” configuration. This is a com- tUM Global Atmosphere component coupled internally to mon choice of scientific options that will be applied to all a JULES Global Land component and externally to the of the Met Office’s global atmospheric MetUM components. NEMO ocean model (Madec, 2008) and the Los Alamos sea A similar approach is planned for other components of the ice (CICE) model (Hunke and Lipscombe, 2008) via the OA- Earth System such as the land surface, ocean, sea-ice and SIS coupler (Valcke, 2006). The Met Office’s short-range atmospheric chemistry and aerosols. The combined Global NWP forecasting system currently uses the MetUM Global Atmosphere is a logical progression from the decision to de- Atmosphere and JULES Global Land initialised via a 4D-Var velop a single atmospheric formulation for use across the data assimilation cycle (Rawlins et al., 2007), a soil moisture Met Office Hadley Centre’s climate models: the HadGEM3 analysis system (Best and Maisey, 2002) and uses fixed sea- (Hadley Centre Global Environmental Model version 3) fam- surface temperature (SST) and sea ice fields from the OSTIA ily. In recognition of this, the atmospheric components of (Operational Sea-surface Temperature and sea Ice Analysis) the two operationally used climate modelling systems pre- system (Donlon et al., 2011). viously known as HadGEM3-r1.1 (Hewitt et al., 2011) and To illustrate the breadth of potential applications, the fol- HadGEM3-r4.0 (Arribas et al., 2011) have been labelled lowing is a list of operational or production systems in which GA1.0 and GA2.0, respectively. the Met Office is using or plans to use Global Atmosphere This paper describes the first such MetUM configuration configurations: developed for use in short-range NWP as well as climate modelling: Global Atmosphere 3.0 (GA3.0). It also de- – Deterministic NWP (6 days): atmosphere/land-only, scribes a configuration of the JULES (Joint UK Land En- started from 4D-Var data assimilation and soil mois- vironment Simulator, Best et al., 2011; Clark et al., 2011) ture, snow and SST analyses. Produces global weather land surface model developed for use with GA3.0; this is la- forecasts and boundary conditions for limited area NWP belled Global Land 3.0 (GL3.0). In the following section models; we describe precisely what we mean by the Global Atmo- – Ensemble prediction system (15 days): 24 members sphere configuration, the vision of how this could be used with stochastic physics parametrizations initialised by the Met Office and the wider MetUM community and our from perturbed deterministic analyses and using per- plans for a more open development process. Again, this is de- sisted SST anomalies (Bowler et al., 2008). Provides signed as a general template that could be followed for other probabilistic global weather forecasts and boundary Earth System components. In Sect. 3 we scientifically de- conditions for the regional ensemble; scribe both GA3.0 and GL3.0, whilst in Sect. 4 we provide a more detailed description of some of the major develop- – Monthly forecasting system (60 days): Lagged ments made since GA2.0 and its accompanying Global Land ensemble (28 members per week) of coupled Me- configuration GL2.0. Although these configurations are be- tUM/JULES/NEMO/CICE models with stochastic ing assessed in NWP trials and case studies, the configura- physics parametrizations initialised from deterministic tions used in the first operational global NWP implementa- atmospheric and ocean model analyses. Output is tion based on GA3.0 and GL3.0 retain a small number of bias-corrected using an extensive set of hindcasts;

Geosci. Model Dev., 4, 919–941, 2011 www.geosci-model-dev.net/4/919/2011/ D. N. Walters et al.: MetUM GA3.0/3.1 and JULES GL3.0/3.1 configurations 921

– Seasonal forecasting system (210 days): Extension of and aerosols. Full Earth System models (e.g., HadGEM2- the monthly system in which each simulation runs for ES, described in Collins et al., 2008) represent the emis- 7 months. A lagged 42-member ensemble is updated sion, transport, radiative impact and chemical interaction of every week by including all forecasts run in the previous atmospheric aerosols and trace gases as well as their im- 21 days (Arribas et al., 2011); pact on other elements of the Earth System such as the bio- sphere. Lower down the hierarchy of complexity are climate – Decadal prediction system (10 yr): using coupled Me- models that include a full suite of prognostic atmospheric tUM/JULES/NEMO/CICE models and based on that aerosols but little atmospheric chemistry, such as HadGEM2- discussed in Smith et al. (2007); AO (HadGEM2 Development Team: Martin et al., 2011), – HadGEM3-A Atmosphere/land-only climate whilst short-range NWP models currently tend not to in- assessments/inter-comparisons (30 yr); clude prognostic aerosol at all. For this reason, we choose to consider atmospheric chemistry and aerosols as a com- – Continent-scale regional climate model ponent of the Earth System separate to the Global Atmo- assessments/inter-comparisons (30 yr) or climate sphere. To maintain traceability through this hierarchy of change projections (up to 150 yr); models, however, we treat their interaction with atmospheric physics schemes as consistently as possible. For example, – HadGEM3-AO MetUM/JULES/NEMO/CICE coupled using too simple an aerosol climatology in NWP models has climate integrations for assessment and climate change previously led to regional or seasonal biases in the radiation projections (100 s of years); budget (Haywood et al., 2005), which have recently been – Planned HadGEM3 Earth System model (100 s of partly overcome by introducing a set of three-dimensional, years). seasonal, speciated aerosol climatologies derived from the prognostic fields of a long climate integration. Outside the Met Office, both the MetUM and JULES are Finally, from a technical stance, the Global Atmosphere is widely utilised within both UK and international academia independent of MetUM code version. Assessments of GA3.0 as well as being used operationally by research institutes and and GA3.1 have been performed at MetUM vn7.7 and vn7.8; meteorological services across the globe. Each user will have because the vast majority of options within the model are their own projects, systems and frameworks in which they backward compatible, future applications may also use code wish to run the models; our aim is to provide technically and vn7.9 and beyond. The use of older code versions is more scientifically robust configurations that our collaborators can complicated, since some of the options in a given GA config- contribute to and use with confidence. uration may not be available in older releases, but for minor To be applicable across this variety of systems, the physi- changes it is possible to introduce branches to the main code cal formulation of the Global Atmosphere must be indepen- trunk that include the newer code1. We only plan to support dent of horizontal resolution. The MetUM contains a few a given Global Atmosphere configuration across a handful of options and input parameters that have an implicit resolu- MetUM code releases, which means that the extended use of tion dependence, so these also remain outside our definition stable configurations for major climate change research ac- of the Global Atmosphere. Ideally, the configuration would tivities is likely to require the continued use of an older code also be independent of vertical resolution, allowing users to base. add refinement to the regions of the atmosphere in which they are particularly interested (e.g. see discussion in Senior et al., 2.2 The Global Atmosphere development process 2010). In practice, however, the performance of parametriza- tions in the troposphere is still sensitive to vertical resolu- Global Atmosphere configurations are developed over an an- tion, so currently we are only assessing GA3.0 in systems nual cycle in which any potential enhancements are tested for with a common vertical tropospheric resolution, which is de- impact across all timescales. These changes are combined scribed in Sect. 3.1. In the stratosphere and above, how- into a small number of test configurations and then a final ever, there is less constraint on the vertical resolution and frozen configuration. This is then evaluated through a full we use different level sets in different applications. Current set of climate assessments and NWP trials including cou- research plans include an investigation into the benefits of pling to other components of the Earth System and including increased vertical resolution in different layers of the atmo- non-production systems such as very high resolution climate sphere across all timescales, which may lead to some refine- simulations and NWP case studies across a whole range of ment in the level sets used in future Global Atmosphere re- 1For example, a small proportion of the code for GA3.0 is not leases. available in the central trunk at vn7.7. Because it is best practice to Another field where the variety of applications requires only use centrally lodged code, wherever possible we limit this to flexibility in the configuration is in the interaction between code reviewed and lodged at a later MetUM version and changes to the atmosphere and other components of the Earth System. “parameter” statements that can be introduced as user set options in One such example is the treatment of atmospheric chemistry a later release. www.geosci-model-dev.net/4/919/2011/ Geosci. Model Dev., 4, 919–941, 2011 922 D. N. Walters et al.: MetUM GA3.0/3.1 and JULES GL3.0/3.1 configurations resolutions. If the configuration performs sufficiently across By convention, global configurations are defined on 2n these tests, it is introduced in subsequent upgrades to the Met longitudes and 1.5n + 1 latitudes of scalar grid-points with Office systems described above and “released” for use by the scalar and zonal wind variables at the north and south poles. wider community. If some changes to the configuration are This choice makes the grid-spacing approximately isotropic required, however, due to insufficient performance in a par- in the mid-latitudes and means that the integer n, which rep- ticular system or set of systems, then this is overcome by the resents the maximum number of zonal 2 grid-point waves introduction of a branch to the Global Atmosphere develop- that can be represented by the model, uniquely defines its ment trunk. An example of this is the GA3.1 configuration horizontal resolution; a model with n = 96 is said to be N96 that the Met Office will be implementing – and recommend- resolution. Limited area configurations use a rotated longi- ing for external use – in NWP systems. This branch con- tude/latitude grid with the pole rotated so that the grid’s equa- figuration, however, is due to a continued small set of long- tor runs through the centre of the model domain. standing differences between our NWP and climate models In the vertical, the majority of climate configurations we rather than deficiencies in any new developments. Although have assessed use the high-top 85 level set discussed in He- we will not be using GA3.0 in operational NWP systems, we witt et al. (2011). We label this L85(50t,35s)85, which refers are assessing its performance in NWP trialling. This is im- to the fact that it has 50 levels below 18 km (and hence at portant not only in understanding how best to resolve these least sometimes in the troposphere), 35 levels above this (and differences over the coming year, but also because we will hence solely in or above the stratosphere) and a fixed model test the performance of each new development against the lid 85 km above the surface. Limited area climate simula- Global Atmosphere trunk and not its branches; this will re- tions use a reduced 63 level set, L63(50t,13s)40, which has duce the probability of additional changes being introduced the same 50 levels below 18 km, with only 13 above and with compensating errors that make it harder to remove the a lower model top at 40 km. Finally, NWP configurations use necessity for a branch. Finally, the Met Office has introduced a 70 level set, L70(50t,20s)80 which has an almost identical an open development process and will encourage the scrutiny 50 levels below 18 km, a model lid at 80 km, but has a re- and participation of MetUM collaborators. We believe that duced stratospheric resolution compared to L85(50t,35s)85. inviting the wider MetUM community to contribute to the Although we have used a range of vertical resolutions in the Global Atmosphere configuration will add to its scientific stratosphere, a consistent tropospheric vertical resolution is integrity, encourage its use and accelerate its improvement. used throughout as noted in Sect. 2. The same argument also stands for the other components of the Earth System. 3.2 Solar and terrestrial radiation In the following sections, we provide scientific descrip- tions of GA3.0 and GL3.0. For brevity, we refer to the cou- Shortwave (SW) radiation from the Sun is absorbed in the pled Global Atmosphere and Global Land configurations as atmosphere and at the Earth’s surface and provides the en- GA3.0/GL3.0. Major developments since GA2.0/GL2.0 are ergy to drive the atmospheric circulation. Longwave (LW) discussed in more detail in Sect. 4. radiation is emitted from the planet into space and also re- distributes heat within the atmosphere. These processes are parametrized via the radiation scheme, which provides prognostic atmospheric temperature increments and surface 3 Global Atmosphere 3.0 and Global Land 3.0 fluxes and additional diagnostic fluxes. The radiation scheme of Edwards and Slingo (1996) is 3.1 Dynamical formulation and discretization used with a configuration based on Cusack et al. (1999) with a number of significant updates. The correlated-k method The MetUM’s dynamical core uses a semi-implicit semi- is used for gaseous absorption with 6 bands in the SW Lagrangian formulation to solve the non-hydrostatic, fully- and 9 bands in the LW. The method of equivalent extinc- compressible deep atmosphere equations of motion (Davies tion (Edwards, 1996) is used for minor gases in each band. et al., 2005). The primary dry atmospheric prognostics are Gaseous absorption coefficients are generated using the HI- the three-dimensional wind components, potential temper- TRAN 2001 spectroscopic database (Rothman et al., 2003) ature, Exner pressure, and density, whilst moist prognos- with updates up to 2003. The water vapour continuum is tics such as specific humidity and prognostic cloud fields as represented using version 2.4 of the Clough–Kneizys–Davies well as other atmospheric loadings are advected as free trac- (CKD) model (Clough et al., 1989; Mlawer et al., 1999). 21 ers. These prognostic fields are discretized horizontally onto k-terms are used for the major gases in the SW bands. Ab- a regular longitude/latitude grid with Arakawa C-grid stag- sorption by water vapour (H2O), ozone (O3), carbon dioxide gering (Arakawa and Lamb, 1977), whilst vertical decompo- (CO2) and oxygen (O2) is included. The treatment of O3 ab- sition is done via Charney-Phillips staggering (Charney and sorption is as described in Zhong et al. (2008). The solar Phillips, 1953) using terrain-following hybrid height coordi- spectrum uses data from Lean (2000) at wavelengths shorter nates. than 735 nm with the Kurucz and Bell (1995) spectrum at

Geosci. Model Dev., 4, 919–941, 2011 www.geosci-model-dev.net/4/919/2011/ D. N. Walters et al.: MetUM GA3.0/3.1 and JULES GL3.0/3.1 configurations 923 longer wavelengths. 47 k-terms are used for the major gases altered from the original Tripoli and Cotton (1980) formu- in the LW bands. Absorption by H2O, O3, CO2, CH4, ni- lation to a liquid content where the number of drops over −3 trous oxide (N2O), CFC-11 (CCl3F), CFC-12 (CCl2F2) and 20 µm is 1000 cm . Both these modifications have been HFC134a (CH2FCF3) is included. For climate simulations, described in Abel et al. (2010). In addition, we have used the atmospheric concentrations of CFC-12 and HFC134a are the fall velocities of Abel and Shipway (2007), which allow adjusted to represent absorption by all the remaining trace a better representation of the drizzle drop spectrum. This halocarbons. The treatment of CO2 and O3 absorption is has been combined with a prognostic rain formulation, which as described in Zhong and Haigh (2000) to provide accurate allows three-dimensional advection of the precipitation par- stratospheric heating. ticles. We also make use of multiple sub-timesteps of the Absorption and scattering by the following categories of precipitation scheme, as in Posselt and Lohmann (2008), to aerosol, either prognostic or climatological, are included in achieve a reduction in surface drizzle rates, with one sub- both the SW and LW: ammonium sulphate, mineral dust, sea- timestep for every two minutes of the model timestep. These salt, biomass-burning, fossil-fuel black carbon, fossil-fuel new modifications are described in more detail in Sect. 4. organic carbon, and secondary organic (biogenic) aerosols. When prognostic aerosols are used, the aerosol mass mixing The parametrization of cloud droplets is described in Ed- ratios provide the cloud droplet number for autoconversion, wards and Slingo (1996) using the method of “thick averag- according to the formulae of Jones et al. (1994). In GA3.0, ing”. Pade´ fits are used for the variation with effective radius, the aerosols which provide the droplet number are sulphur, which is computed from the number of cloud droplets. When soot, biomass and fossil fuels/organic carbon. When using using prognostic aerosol, this is derived from the aerosol either no aerosol or climatological aerosol, the cloud droplet concentrations (Jones et al., 2001), whilst when using either number assumes the same land/sea split as in the radiation no aerosol or climatological aerosol, this is assumed to be scheme. 100 cm−3 for maritime airmasses and 300 cm−3 for conti- nental airmasses. The parametrization of ice crystals is de- 3.4 Large scale cloud scribed in Edwards et al. (2007). Full treatment of scattering is used in both the SW and LW. The sub-grid cloud structure Clouds appear on sub-grid scales well before the humidity is represented using the Monte-Carlo Independent Column averaged over the size of a model grid-box reaches satura- Approximation (McICA) as described in Hill et al. (2011), tion. A cloud parametrization scheme is therefore required with optimal sampling using 6 extra terms in the LW and 10 to determine the fraction of the grid-box which is covered in the SW for the reduction of random noise. by cloud and the amount and phase of condensed water con- Full radiation calculations are made every 3 h using the in- tained in those clouds. The formation of clouds will con- stantaneous cloud fields and a mean solar zenith angle for vert water vapour into liquid or ice and release latent heat. the following 3 h period. Corrections for the change in so- The cloud cover and liquid and ice water contents are then lar zenith angle on every model timestep and the change in used by the radiation scheme to calculate the radiative im- cloud fields every hour are made as described in Manners pact of the clouds and by the microphysics scheme to cal- et al. (2009). The land surface is prescribed a global emissiv- culate whether any precipitation has formed. GA3.0 uses ity of 0.97 whilst the albedo is set by the land surface model. the MetUM’s prognostic cloud fraction and prognostic con- The direct SW flux at the surface is corrected for the angle densate (PC2) scheme (Wilson et al., 2008a,b). This uses and aspect of the topographic slope as described in Manners three prognostic variables for humidity mixing ratio: water (2011). The albedo of the sea surface uses a modified ver- vapour, liquid and ice and a further three prognostic vari- sion of the parametrization from Barker and Li (1995) with ables for cloud fraction: liquid, ice and total. The total a varying spectral dependence. cloud fraction is not necessarily equal to the sum of the other two due to the presence of mixed-phase regions. The fol- 3.3 Large scale precipitation lowing atmospheric processes can modify the cloud fields: shortwave radiation, longwave radiation, boundary layer pro- The formation and evolution of precipitation due to grid scale cesses, convection, precipitation, small scale mixing, advec- processes is the responsibility of the large scale precipitation tion and pressure changes due to large scale vertical motion. – or microphysics – scheme, whilst small scale precipitating The convection scheme calculates increments to the prognos- events are handled by the convection scheme. The micro- tic liquid and ice water contents by detraining condensate physics scheme has prognostic input fields of heat and mois- from the convective plume, whilst the cloud fractions are up- ture, which it modifies in turn. The microphysics used in dated using the non-uniform forcing method of Bushell et al. GA3.0 is based on Wilson and Ballard (1999), with exten- (2003). One advantage of the prognostic approach is that sive modifications. The particle-size distribution has been clouds can be transported away from where they were cre- modified from a Marshall and Palmer (1948) distribution to ated. For example, anvils detrained from convection can per- include an intercept based on rain rate and the minimum sist and be advected downstream long after the convection cloud liquid content for autoconversion to occur has been itself has ceased. Additionally, cloud budget tendency terms www.geosci-model-dev.net/4/919/2011/ Geosci. Model Dev., 4, 919–941, 2011 924 D. N. Walters et al.: MetUM GA3.0/3.1 and JULES GL3.0/3.1 configurations from the prognostic cloud scheme can be analysed to deter- represent a realistic quasi-biennial oscillation in the tropics. mine how each of the physical processes in the model is con- The scheme, based on Warner and McIntyre (2001), treats tributing to the evolution of the cloud fields (Morcrette and a spectrum (against vertical wavenumber) of gravity waves Petch, 2010). in 4 azimuthal directions (W, N, E, S) and represents the pro- cesses of wave generation, conservative propagation and dis- 3.5 Orographic gravity wave drag sipation by critical-level filtering and wave saturation. The simplest wave generation (Scaife et al., 2002) is to launch, The effect of local and mesoscale orographic features (in- in the lower troposphere, a vertical flux of horizontal grav- dividual hills through to small mountain ranges) not re- ity wave momentum that totals a globally invariant 2.47 mPa solved by the mean orography must be parametrized. The in each direction. This represents order 10 % of the satura- smallest scales, where buoyancy effects are not important, tion spectrum at launch height. Isotropic fluxes guarantee are represented by an effective roughness parametrization zero total momentum at launch height and momentum con- in which the roughness length for momentum used by the servation is also enforced at the model upper boundary by boundary layer scheme is increased over orography (Gre- imposing an “opaque lid”, i.e. a condition of zero vertical gory et al., 1998). The effects of the remainder of the sub- wave flux passing up out of the top layer. The launch spec- grid orography (on scales where buoyancy effects are impor- trum has the two-part form of Warner and McIntyre (2001) tant) are parametrized by a flow blocking and gravity wave with low wavenumber cut-off and spectrum peak located at drag parametrization. The scheme is based on Webster et al. wavelengths 20 km and 4.3 km, respectively. It is linear for (2003) and accounts for drag effects due to sub-grid orogra- wavenumbers up to the peak, beyond which it has the inverse phy in stable conditions. The sub-grid orography is described cubic tail characteristic of saturation. The launched spec- in terms of its amplitude, which is proportional to the stan- tra propagate upwards conservatively, responding to varia- dard deviation of the source orography in a model grid-box, tions in mean wind speed (Doppler shift) and reducing den- and its anisotropy, i.e. how ridge-like the sub-grid orography sity, under the assumed mid-frequency approximation to the is. The total surface stress is proportional to the bulk low- dispersion equation. Hence wave reflection is not modelled, level winds and stability, and is determined using a simple though its effect is partly accounted for by a reduced launch linear hydrostatic expression; idealised modelling (e.g. Wells spectrum amplitude. Momentum deposition occurs as the et al., 2005) suggests this captures the total surface stress rea- integrated flux reduces due to erosion of each transformed sonably well. The low-level Froude number is then used to launch spectrum that exceeds the locally evaluated satura- partition the total stress into gravity wave and flow block- tion spectrum. As applied, the calculation of integrated flux ing components due to flow over and around the orography, filtering through individual levels first modifies the original respectively. The flow blocking drag is diagnosed assum- launch spectrum to limit flux propagating through the level ing a linear decrease in the stress over the depth of the sub- to below local saturation, then maintains a form of causal- grid orography, whilst the gravity wave stress is launched ity by requiring that integrated flux leaving through the top upwards and a drag exerted at levels where wave breaking of a layer never exceeds the integrated flux which entered or wave saturation is diagnosed. Typically, in excess of 90 % through the layer base. of the global mean of the total surface stress is attributed to low-level flow blocking. The drag is applied as explicit in- 3.7 Atmospheric boundary layer crements to the model wind fields, so a numerical limiter is imposed on the flow blocking drag to ensure the numerical Turbulent motions in the atmosphere are not resolved by stability of the scheme (Brown and Webster, 2004). global atmospheric models, but are important to parametrize in order to give realistic vertical structure in the thermody- 3.6 Non-orographic gravity wave drag namic and wind profiles. Although referred to as the “bound- ary layer” scheme, this parametrization represents mixing Orography is just one forcing mechanism for gravity waves; over the full depth of the troposphere. The scheme is that others (e.g. convection, fronts, jets) can force gravity waves of Lock et al. (2000) with the modifications described in with non-zero phase-speed. The temporal and spatial resolu- Lock (2001) and Brown et al. (2008). It is a first-order turbu- tion of a given model configuration dictate the scale of grav- lence closure mixing adiabatically conserved variables. For ity waves sustained explicitly; accelerations by the remainder unstable boundary layers diffusion coefficients (K-profiles) are generated by a sub-grid parametrization scheme (Scaife are specified functions of height within the boundary layer, et al., 2002). It is important to represent momentum depo- related to the strength of the turbulence forcing. Two sep- sition by breaking of such waves in the upper stratosphere arate K-profiles are used, one for surface sources of turbu- and mesosphere because they drive a circulation that op- lence (surface heating and wind shear) and one for cloud- poses radiative equilibrium in zonal mean wind and temper- top sources (radiative and evaporative cooling). The exis- ature structure at these heights. Also, models without such tence and depth of unstable layers is diagnosed initially by schemes report a need for much finer vertical resolution to moist adiabatic parcels and then adjusted to ensure that the

Geosci. Model Dev., 4, 919–941, 2011 www.geosci-model-dev.net/4/919/2011/ D. N. Walters et al.: MetUM GA3.0/3.1 and JULES GL3.0/3.1 configurations 925 buoyancy consumption of turbulence kinetic energy is lim- The deep convection scheme differs from the original Gre- ited. This can permit the cloud layer to decouple from the gory and Rowntree (1990) scheme in using a convective surface (Nicholls, 1984). If cumulus convection is diagnosed available potential energy (CAPE) closure based on Fritsch (through comparison of cloud and sub-cloud layer moisture and Chappell (1980). Mixing detrainment rates now depend gradients), the surface-driven K-profile is restricted to be- on relative humidity and forced detrainment rates adapt to the low cloud base and the mass flux convection scheme is trig- buoyancy of the convective plume (Derbyshire et al., 2011). gered from that level. Mixing across the top of the boundary The CMT scheme uses a flux gradient approach (Stratton layer is through an explicit entrainment parametrization that et al., 2009). is coupled to the radiative fluxes and the dynamics through The shallow convection scheme uses a closure based on a sub-grid inversion diagnosis. If the thermodynamic con- Grant (2001) and has larger entrainment rates than the deep ditions are right, cumulus penetration into a stratocumu- scheme consistent with cloud resolving model (CRM) sim- lus layer can generate additional turbulence and cloud-top ulations of shallow convection. The shallow CMT uses entrainment in the stratocumulus by enhancing evaporative flux-gradient relationships derived from CRM simulations of cooling at cloud-top. There are additional non-local fluxes shallow convection (Grant and Brown, 1999). of heat and momentum in order to generate more vertically The mid-level scheme operates on any instabilities found uniform potential temperature and wind profiles in convec- in a column above the top of deep or shallow convection or tive boundary layers. For stable boundary layers and in the above the lifting condensation level. The scheme is largely free troposphere a local Richardson number scheme (Smith, unchanged from Gregory and Rowntree (1990), but uses the 1990) is used with the stable stability dependence given over Gregory et al. (1997) CMT scheme and a CAPE closure. the sea by the “sharp” function and over land by the “MES- The mid-level scheme operates mainly either overnight over tail” function (which matches linearly between an enhanced land when convection from the stable boundary layer is no mixing function at the surface and “sharp” at 200 m and longer possible or in the region of mid-latitude storms. Other above). This additional near-surface mixing is motivated cases of mid-level convection tend to remove instabilities by the effects of surface heterogeneity, such as those de- over a few levels and do not produce much precipitation. scribed in McCabe and Brown (2007). The resulting diffu- The timescale for the CAPE closure, which is used for the sion equation is solved implicitly using the monotonically- deep and mid-level convection schemes, is essentially fixed damping, second-order-accurate, unconditionally-stable nu- at a chosen value; however, if extremely high large scale ver- merical scheme of Wood et al. (2007). The kinetic energy tical velocities are detected in the column then the timescale dissipated through the turbulent shear stresses is returned to is rapidly reduced to ensure numerical stability. The choice the atmosphere as a local heating term. of timescale for the CAPE closure depends on a pragmatic 3.8 Convection balance between model stability and model performance. In practice, this means that the timescale is currently dependent The convection scheme represents the sub-grid scale trans- on the horizontal resolution and hence it falls outside the def- port of heat, moisture and momentum associated with cumu- inition of GA3.0. lus clouds within a grid-box. The MetUM uses a mass flux convection scheme based on Gregory and Rowntree (1990) 3.9 Structure of the atmospheric model timestep with various extensions to include down-draughts (Gre- gory and Allen, 1991) and convective momentum transport The order of the physical parametrizations described above (CMT). The current scheme consists of three stages: (i) con- within the model timestep and their coupling to the atmo- vective diagnosis to determine whether convection is possi- spheric model’s dynamics can be considered part of the de- ble from the boundary layer; (ii) a call to the shallow or deep sign of the dynamical core. This requires a considered bal- convection scheme for all points diagnosed deep or shal- ance between stability, computational cost and both physical low by the first step; (iii) a call to the mid-level convection and numerical accuracy. The MetUM’s timestepping treats scheme for all grid-points. slow timescale processes in parallel prior to the main ad- The convective diagnosis is based on an undilute parcel vection step. This is followed by the fast timescale pro- ascent from the near surface for grid-boxes where the sur- cesses, which are treated sequentially, prior to the final dy- face layer is unstable and forms part of the boundary layer namical solution (Staniforth et al., 2002). In this frame- diagnosis (Lock et al., 2000). Shallow convection is diag- work, the slow processes include radiation, large scale pre- nosed if the following conditions are met: (i) the parcel at- cipitation and gravity wave drag, whilst the fast processes tains neutral buoyancy below 2.5 km or below the freezing include atmospheric (boundary layer) turbulence, convection level whichever is higher; (ii) the air in model levels forming and coupling to the land surface model. Prognostic cloud a layer of order 1500 m above this has a mean vertical ve- variables and tracers such as aerosols are advected by the −1 locity less than 0.02 m s . Otherwise, convection diagnosed semi-Lagrangian dynamics. Their sources and sinks oc- from the boundary layer is defined as deep. cur where appropriate within the physical parametrization schemes (e.g. phase changes in the microphysics, wash-out www.geosci-model-dev.net/4/919/2011/ Geosci. Model Dev., 4, 919–941, 2011 926 D. N. Walters et al.: MetUM GA3.0/3.1 and JULES GL3.0/3.1 configurations in large scale precipitation and mixing in the boundary layer penetration of light within the vegetation canopy and its sub- and turbulence schemes). Chemical and physical process in- sequent impact on photosynthesis (Mercado et al., 2007). In ternal to the aerosol scheme occur at the end of the timestep. addition, the canopy can interact with snow, with some of the snow intercepted by the canopy itself and the remainder 3.10 Atmospheric aerosols and chemistry being held under the canopy. This impacts on the surface albedo, the snow sublimation and the snow melt. The vege- As discussed in Sect. 2, the modelling of atmospheric tation canopy code has been adapted for use with the urban aerosols and chemistry is considered as a separate compo- surface type by defining an “urban canopy” with the thermal nent of the full Earth System and remains outside the scope properties of concrete (Best, 2005). This has been demon- of this document. The impact of aerosols on atmospheric strated to give improvements over representing an urban area parametrizations, however, is part of the Global Atmosphere as a rough bare soil surface. Similarly, this canopy approach component and has therefore been included in the descrip- has also been adopted for the representation of lakes. The tions above. The treatment of tropospheric aerosols in sys- original representation in MOSES was through a soil surface tems which do not model these explicitly is supplemented by that could evaporate at the potential rate (i.e. a soggy soil), the use of a three-dimensional monthly climatology for each which can been shown to have incorrect seasonal and diurnal aerosol species, although currently these are only used to cycles for the surface temperature. By defining an “inland model the direct aerosol effect. In addition to the treatment of water canopy” and setting the thermal characteristics to those these tropospheric aerosols, GA3.0 includes a simple strato- of a suitable depth of water (taken to be 1 m), a better diur- spheric aerosol climatology based on Cusack et al. (1998); nal cycle for the surface temperature is achieved. For Earth it also includes the production of stratospheric water vapour System modelling JULES includes soil carbon, comprising via a simple methane oxidation parametrization (Untch and of four carbon pools, which is based on the RothC soil car- Simmons, 1999). bon scheme (Jenkinson et al., 1990; Coleman and Jenkinson, 1999). Also included are interactive methane emissions from 3.11 Land surface and hydrology: Global Land 3.0 wetland areas (Gedney et al., 2004). Surface fluxes are calculated separately on each tile using The exchange of fluxes between the land surface and the at- surface similarity theory. In stable conditions the similar- mosphere is an important mechanism for heating and moist- ity functions of Beljaars and Holtslag (1991) are adopted, ening the atmospheric boundary layer. In addition, the ex- whilst in unstable conditions the functions are taken from change of CO2 and other greenhouse gases plays a significant Dyer and Hicks (1970). The effects on surface exchange of role in understanding future climate change. The hydrolog- both boundary layer gustiness (Godfrey and Beljaars, 1991) ical state of the land surface contributes to impacts such as and deep convective gustiness (Redelsperger et al., 2000) are flooding and drought as well as providing fresh water fluxes included. 1.5 m temperatures and 10 m winds are interpo- to the ocean, which influences ocean circulation. Therefore, lated between the model’s grid-levels using the same simi- a land surface model needs to be able to represent this wide larity functions, but a parametrization of transitional decou- range of processes over all surface types that are present on pling in very light winds is included in the calculation of the the Earth. 1.5 m temperature. GL3.0 uses a new community land surface model, JULES Soil processes are represented using a 4 layer scheme for (Joint UK Land Environment Simulator Best et al., 2011; the heat and water fluxes with hydraulic relationships taken Clark et al., 2011), to model all of the processes at the land from van Genuchten (1980). As in MOSES, these four soil surface and in the sub-surface soil. JULES is based on layers have thicknesses from the top down of 0.1, 0.25, 0.65 a combination of the Met Office Surface Exchange Scheme and 2.0 m. The impact of moisture on the thermal charac- (MOSES, Cox et al., 1999) and the TRIFFID (Top-down teristics of the soil is represented using a simplification of Representation of Interactive Foliage and Flora Including Johansen (1975), as described in Dharssi et al. (2009). The Dynamics) dynamic vegetation model (Cox et al., 2000; Cox, energetics of water movement within the soil is taken into ac- 2001). A tile approach is used to represent sub-grid scale count, as is the latent heat exchange resulting from the phase heterogeneity (Essery et al., 2003), with the surface of each change of soil water from liquid to solid states. Sub-grid land point subdivided into five types of vegetation (broadleaf scale heterogeneity of soil moisture is represented using the trees, needleleaf trees, temperate C3 grass, tropical C4 grass Large Scale Hydrology approach (Gedney and Cox, 2003), and shrubs) and four non-vegetated surface types (urban ar- which is based on the topography based rainfall-runoff model eas, inland water, bare soil and land ice). TOPMODEL (Beven and Kirkby, 1979). This enables the Vegetation canopies are represented in the surface en- representation of an interactive water table within the soil ergy balance through the coupling to the underlying soil. that can be used to represent wetland areas, as well as in- This canopy is coupled via radiative and turbulent exchange, creasing surface runoff through heterogeneity in soil mois- whilst any bare soil component couples through conduction. ture driven by topography. JULES also uses a canopy radiation scheme to represent the

Geosci. Model Dev., 4, 919–941, 2011 www.geosci-model-dev.net/4/919/2011/ D. N. Walters et al.: MetUM GA3.0/3.1 and JULES GL3.0/3.1 configurations 927

A river routing scheme is used to route the total runoff from inland grid-points both out to the sea and to inland basins, where it can flow back into the soil moisture. Ex- cess water in inland basins is distributed evenly across all sea outflow points. In coupled model simulations the re- sulting freshwater outflow is passed to the ocean, where it is an important component of the thermohaline circulation, whilst in atmosphere/land-only simulations this ocean out- flow is purely diagnostic. River routing calculations are per- formed using the TRIP (Total Runoff Integrating Pathways) model (Oki and Sud, 1998), which uses a simple advection method (Oki, 1997) to route total runoff along prescribed river channels on a 1 × 1 ◦grid using a 3 h timestep. Land surface runoff accumulated over this timestep is mapped onto the river routing grid prior to the TRIP calculations, after which soil moisture increments and total outflow at Fig. 1. Raindrop fall speed relations as a function of diameter. river mouths are mapped back to the atmospheric grid (Fal- The blue line shows the fall speed relation of Sachidananda and loon and Betts, 2006). This river routing model is not cur- Zrnic´ (1986), which was used in GA2.0 and before. The black rently being used in limited area or NWP implementations line is the observations of Beard (1976) and the red line is the new of GA3.0/GL3.0. parametrization of Abel and Shipway (2007).

4 Developments since Global Atmosphere/Land 2.0 dent Column Approximation (Pincus et al., 2003) which uses In this section, we describe in more detail some of the ma- a different sub-grid cloud profile for each monochromatic ra- jor developments in the Global Atmosphere and Global Land diative transfer calculation. For spectral regions with low configurations since GA2.0/GL2.0 (Arribas et al., 2011). gaseous absorption, where the cloud radiative effect is great- est, extra cloud profiles are sampled in order to reduce ran- 4.1 Solar and terrestrial radiation dom noise. The number of monochromatic calculations are increased by 6 in the LW and 10 in the SW using the method In GA3.0, the treatment of ozone absorption in the ultra- of optimal sampling described in Hill et al. (2011). Inclu- violet has been upgraded based on Zhong et al. (2008). The sion of the radiative effects of cloud inhomogeneity corrects first SW band is divided into six sub-bands each of which has a long-standing bias in the transmissivity of grid-box mean one ozone absorption coefficient. This allows the solar irra- cloud, and has an impact on the top-of-atmosphere radiation diance to be specified for each sub-band facilitating climate balance. experiments which use a varying solar spectral irradiance. The total number of k-terms has increased from 20 in GA2.0 4.2 Large scale precipitation to 21 for the major gases in the SW bands. GA3.0 includes a number of modifications to the micro- The solar spectrum has been upgraded from the Kurucz physics scheme, which were introduced to reduce the spu- and Bell (1995) spectrum used in GA2.0. Below 735 nm, rious occurrence of drizzle that has been made worse by the new spectrum is based on satellite observations meaned the introduction of the PC2 cloud scheme (Wilson et al., over the two solar cycles between 1983 and 2004 inclusive 2008a). The rain fall speeds have been modified from the (data updated from Lean, 2000). Above 735 nm the Kurucz simple power-law relation described by Sachidananda and spectrum is retained. The new spectrum significantly reduces Zrnic´ (1986) to a more complex relation of Abel and Ship- a warm bias that is present in the stratosphere in GA2.0. way (2007) A treatment of cloud inhomogeneity has been introduced d1R −h1RD d2R −h2RD as described in Hill et al. (2011). A stochastic cloud gen- VR(D) = c1RD e +c2RD e , (1) erator based on Rais¨ anen¨ et al. (2004) is used to generate 64 cloud profiles to represent each sub-grid field. This as- where VR is the raindrop fall speed, D is its diameter and (1−d ) −1 sumes the horizontal variation in the in-cloud water con- the constants c1R = 4845.1 m 1R s , d1R = 1.0, h1R = −1 (1−d ) −1 tent follows a Gamma distribution with a fractional standard 195.0 m , c2R = −446.009 m 2R s , d2R = 0.782127 −1 deviation of 0.75. Vertical overlap of the sub-grid cloud and h2R = 4085.35 m . Figure 1 shows the fall speed as assumes “exponential-random-overlap” (Hogan and Illing- a function of raindrop diameter for the Abel and Shipway worth, 2000) with a decorrelation scale of 100 hPa for the (2007) parametrization, the observations of Beard (1976) and cloud fraction and 50 hPa for the condensate. The sub-grid the GA2.0 fall speed relation of Sachidananda and Zrnic´ cloud field is then sampled using the Monte Carlo Indepen- (1986). It can be seen that the Abel and Shipway (2007) www.geosci-model-dev.net/4/919/2011/ Geosci. Model Dev., 4, 919–941, 2011 928 D. N. Walters et al.: MetUM GA3.0/3.1 and JULES GL3.0/3.1 configurations relation provides a closer fit to the Beard (1976) observa- 2011). For this initial implementation, JULES has been tions than the Sachidananda and Zrnic´ (1986) relation. For configured to have as few scientific differences from the drizzle-sized drops of a few microns, the Sachidananda and MOSES-II model used in GL2.0 as possible. The main resid- Zrnic´ (1986) fall speed is up to a factor of ten too high. Using ual difference is in the linearisation assumptions used for the Abel and Shipway (2007) means that we can allow drizzle to surface energy balance. In MOSES, the terms are linearised fall more slowly and evaporate more readily in the sub-cloud around the top soil layer temperature, whilst in JULES they layer. In addition, the accretion rate given by Wilson and are linearised around the previous timestep value of the sur- Ballard (1999) has a fall velocity component; by allowing face temperature. The more accurate linearisation of this a slower fall velocity for smaller drizzle drops in the cloud, term in the energy balance in JULES leads to increased out- the accretion rates are much lower and less drizzle will fall going longwave radiation from the surface and hence a slight out of the cloud base. surface cooling. The most significant advantage of imple- The diagnostic rain scheme used in Wilson and Ballard menting JULES, however, is that its additional functional- (1999) and the GA2.0 configuration makes the assumption ity and flexibility will allow major new developments in the that all rain will fall out of the column in one timestep. How- Global Land configuration in the coming years. For instance, ever, drizzle drops falling at the order of 10 cm s−1 or less JULES currently consists of options for new snow and ur- from a cloud base at around 500 m altitude will take over ban schemes, whilst new modules for plant and soil nitrogen, 5000 s to reach the surface, which is longer than the timesteps crops and an alternative dynamic vegetation scheme are be- used by the systems that follow GA3.0. Thus the diagnos- ing developed. tic rain assumption is not valid. Furthermore, as it is not In coupled systems, GL3.0 introduces the facility to read possible for the diagnostic rain assumption to hold rain in in an iceberg calving ancillary, which is calibrated from the column between timesteps, we have progressed to using a previous model run. This allows a flux of freshwater and a prognostic rain formulation. This has been used opera- associated latent heat to be added to the total flux over the tionally in convection permitting configurations of the Me- ocean with a spatial extent representing the calving of ice- tUM for some time and has been described in detail by Lean bergs from glaciers. This has no impact in atmosphere/land- (2002), who found that in high-resolution simulations, the only systems but, in the absence of a full treatment of land location of rainfall in the vicinity of orography agreed much ice processes, such a term allows ocean models coupled to better with observations when using a prognostic scheme. GA3.0/GL3.0 to remain in freshwater balance. For a more accurate representation of drizzle processes, Pos- selt and Lohmann (2008) show that it is necessary to include a number of sub-timesteps of the microphysics scheme and 5 Global Atmosphere 3.1 and Global Land 3.1 they suggested using 30 sub-timesteps with a 15 min model timestep (i.e. one sub-timestep for each 30 s of the model GA3.0 and GL3.0 are the Met Office’s first attempt at defin- timestep). However, in our experiments, we find that using ing a coupled global atmosphere/land configuration that can one sub-timestep for every 2 min of the timestep provides be used across all timescales. During their development, a very similar results to using more iterations, yet without the number of differences between the global model configura- additional computational cost. tions used in the operational NWP suite and the latest climate configurations were removed, ranging from small changes in 4.3 Convection individual parameters to the introduction of new parametriza- tions. For a small number of these differences, the op- At GA3.0 two of the conditions for the diagnosis of shal- erational NWP suite had not previously been brought into low convection were relaxed, in particular: (i) the require- line with climate configurations, despite knowledge that their ment that an inversion was detected above the level of neu- parametrizations had a more sound physical basis, because tral buoyancy was removed, and (ii) the maximum large scale these changes had small but detrimental impacts on NWP vertical velocity allowed above the level of neutral buoyancy performance. In order to maximise the possibility of imple- was increased from 0.0 m s−1 to 0.02 m s−1. These changes menting a Global Atmosphere based configuration in the op- effectively make the conditions for the diagnosis of deep con- erational NWP suite, therefore, we took the conservative ap- vection more stringent. They were made to prevent the spuri- proach of re-introducing a small number of these differences ous diagnosis of deep convection in situations where the deep by defining the branch configurations Global Atmosphere 3.1 convection scheme was not designed to operate. (GA3.1) and Global Land 3.1 (GL3.1); these branch config- urations were tested for inclusion in the NWP suite ahead 4.4 Land surface and hydrology of GA3.0/GL3.0. Global NWP tests of GA3.0/GL3.0 do show reasonable performance, but lead to a small reduction In the Global Land configuration, the major change with in forecast skill for several important parameters. Whilst the GL3.0 is the replacement of the MOSES-II (Essery et al., operational implementation of GA3.1/GL3.1 means that we 2003) model with JULES (Best et al., 2011; Clark et al., have not yet adopted a truly seamless forecast system, it will

Geosci. Model Dev., 4, 919–941, 2011 www.geosci-model-dev.net/4/919/2011/ D. N. Walters et al.: MetUM GA3.0/3.1 and JULES GL3.0/3.1 configurations 929 allow us time to understand why this performs better than GL3.1 is not able to hold separate snow stores above and GA3.0/GL3.0 at short forecast ranges; we aim to remove below the tree canopy. To compensate for this, GL3.1 uses as many of these differences as possible during the devel- a snow-covered albedo for trees of 0.15, rather than the 0.25 opment of GA4.0/GL4.0. Finally, the development process used in GL3.0. The amalgamated tile approach also means itself relies on using GA3.0/GL3.0 and not GA3.1/GL3.1 as that in GL3.1 we cannot introduce an “inland water canopy” its future baseline; this will stop the introduction of addi- to mimic the increased thermal capacity of deep lakes. The tional compensating errors that make it harder to move away other land surface differences are that GL3.1 uses a bare soil from the settings in the branch configurations. roughness length of 3.2 mm and a sea-ice roughness length The primary differences between GA3.1 and GA3.0 are of 100 mm for marginal ice zones and 3 mm for pack ice. in the radiation and boundary layer schemes, whilst GL3.1 In GL3.0 these values are 0.3 mm, 0.5 mm and 0.5 mm, re- and GL3.0 differ most significantly in their treatment of the spectively. GL3.1 does not use the Large Scale Hydrology land surface tiles and the runoff of precipitation from the soil. approach to calculate its surface and sub-surface runoff, and Each of these differences is described in detail below. therefore has no representation of sub-grid heterogeneity in this process. 5.1 Solar and terrestrial radiation

With a focus on computational speed and improved tropo- 6 Preliminary model evaluation spheric heating rates, GA3.1 does not include the enhanced treatment of CO2 and O3 LW absorption (Zhong and Haigh, An important part of the MetUM Global Atmosphere devel- 2000). This reduces the number of k-terms from 47 to 33 for opment process is its evaluation across a variety of resolu- the major gases in the LW bands, with a similar reduction in tions and timescales. The principle of this evaluation follows the computation time. The optical properties of ammonium that previously used for assessing the performance of the Met nitrate are included as a “delta” aerosol climatology to repre- Office’s climate and NWP models in which the the model sent a missing process left by the other aerosol climatologies is assessed using a relatively large basket of quantitative when compared to observations. measures, which are collated and investigated qualitatively. At the time of writing, the evaluation of the GA3.0/GL3.0 5.2 Atmospheric boundary layer configurations in a large subset of the systems listed in Sect. 2.1 is underway. A full assessment of the performance As discussed in Brown et al. (2008), the Met Office’s global of GA3.0/GL3.0 is beyond the scope of this paper. In this NWP model has for many years used stable stability func- section, however, we present a small set of results from some tions, together with adjustments to the mixing lengths, that of the evaluation simulations that have already completed. combine to give significantly more mixing than the GA3.0 setup and this remains as a difference in GA3.1. This is, in 6.1 N96 HadGEM3 30 yr AMIP climate simulation fact, common practice in many NWP centres and typically has been found to reduce spurious cooling at the surface and Because this is only the first in a planned series of papers, increase skill in predicting the synoptic evolution (Viterbo there are no results yet published from an equivalent as- et al., 1999; Beljaars and Viterbo, 1998). Although the argu- sessment of GA2.0/GL2.0. Instead, we present a compar- ment can be made that these functions are parametrizing the ison of GA3.0/GL3.0 in a 30 yr atmosphere/land-only cli- effects of heterogeneity, studies suggest this enhancement is mate integration at N96 resolution (approximately 135 km excessive (e.g. see discussion in McCabe and Brown, 2007) in the mid-latitudes) with an equivalent N96 simulation of and that they may simply be compensating for errors else- HadGEM2-A (HadGEM2 Development Team: Martin et al., where in the surface energy balance. 2011). Both GA3.0/GL3.0 and HadGEM2-A simulations 5.3 Convection use the same sea surface temperatures, sea ice fractions, Unlike GA3.0, GA3.1 uses a CAPE closure timescale that CO2 concentrations and other external forcings as the At- is independent of horizontal resolution and which is set to mospheric Model Intercomparison Project (AMIP) frame- 30 min. All other convection options are identical between work of the Coupled Model Intercomparison Project phase GA3.0 and GA3.1. 5 (CMIP5, Taylor et al., 2009). Both simulations cover the period from 1979 to 2008. 5.4 Land surface and hydrology: Global Land 3.1 The overall assessment of the global tropospheric circu- The most significant difference between GL3.1 and GL3.0 lation requires a comparison of the model’s mean flow and is that GL3.1 does not perform its land surface calculations modes of variability against a collection of trusted obser- on each of the 9 surface tiles separately, but amalgamates vational climatologies. A concise way of presenting these the properties of each surface type weighted by their grid- results is through a set of normalised assessment criteria, box fraction into a single tile. As well as being only an ap- each of which are calculated as the root mean square error proximation of a truly tiled surface, this choice means that of a meaned field in GA3.0/GL3.0 divided by the root mean www.geosci-model-dev.net/4/919/2011/ Geosci. Model Dev., 4, 919–941, 2011 930 D. N. Walters et al.: MetUM GA3.0/3.1 and JULES GL3.0/3.1 configurations

2.0

1.5 HadGEM3-A GA3.0 N320 HadGEM3-A GA3.0 N216 HadGEM3-A GA3.0 N96 1.0 HadGEM2-A N96

Normalised assessment criteria 0.5

0.0 PMSLPMSL DJFPMSL MAMglobalPMSL JJA Precipitationglobal SONglobalPrecipitation globalPrecipitation PrecipitationDJF PrecipitationMAMtropical_land PrecipitationJJA tropical_land PrecipitationSONtropical_land PrecipitationDJF tropical_land PrecipitationMAMtropical_ocean PrecipitationJJA tropical_ocean PrecipitationSONtropical_ocean PrecipitationDJF tropical_ocean PrecipitationMAMnorth PrecipitationJJA north PrecipitationSONnorth PrecipitationDJF north 200MAMsouth hPa200JJA south horizhPa850SONsouth hPahoriz850 windsouth hPahorizZ-mean wind DJF horizZ-mean wind JJAglobal UZ-mean wind windDJFglobal UZ-mean wind JJA globalDJF VZ-mean windglobal JJAglobal VZ-mean wind DJFglobal meridPrecipitable JJAglobal meridPrecipitable streamfunction globalZ-mean streamfunction waterZ-mean waterrelativeZ-mean DJF DJFrelativeZ-mean JJAglobal humidity JJAairglobal1.5m global temperaturehumidity airglobal1.5m temperature temperature DJF temperature JJAglobal DJF global DJF JJAglobal JJAglobal global global

Fig. 2. Normalised assessment criteria (ratios of mean field root mean square errors) for a range of atmospheric fields from GA3.0/GL3.0 HadGEM3-A simulations across a range of resolutions compared to an N96 HadGEM2-A baseline. Statistics shown are from the seasons December to February (DJF), March to May (MAM), June to August (JJA) and September to November (SON) and for regions global, tropical land (land points between 30◦ N and 30◦ S), tropical ocean (ocean points between 30◦ N and 30◦ S), north (30◦–90◦ N) and south (30◦–90◦ S). The observation datasets used are HadSLP2 pressure at mean sea level (Allan and Ansell, 2006), GPCP precipitation (Adler et al., 2003), SSMI precipitable water (Wentz and Spencer, 1998) and CRUTEM3 1.5 m temperature (Brohan et al., 2006), whilst the remaining climatologies are from ERA-interim reanalyses (Berrisford et al., 2009). The whisker bars are observational uncertainty, which is calculated by comparing these with alternative datasets; these are ERA-40 pressure at mean sea level and precipitable water (Uppala et al., 2005), CMAP precipitation (Xie and Arkin, 1997), Legates and Willmott (1990) 1.5 m temperature and MERRA reanalyses for everything else (Bosilovich, 2008). The grey shading is the uncertainty associated with internal climate variability, which is estimated by comparing various 20 yr segments of a long coupled simulation. The points represent simulations of GA3.0/GL3.0 at N96, N216 and N320 resolution, which ran for 30, 30 and 20 yr, respectively. square error of the same field in HadGEM2-A. The triangle outside the range of uncertainty in the observations (orange) symbols in Fig. 2 show the normalised assessment criteria whilst 15 lie within this range (green). These results indi- from this N96 AMIP simulation for a range of atmospheric cate that the performance of GA3.0/GL3.0 is a considerable fields, namely: pressure at mean sea level, precipitation, hor- improvement over that of HadGEM2-A; a similar analysis izontal winds at 200 and 850 hPa, precipitable water and of top-of-atmosphere and surface radiation fluxes also shows screen-level temperature as well as zonal means of zonal and a similar improvement. meridional winds, meridional streamfunction, temperature and relative humidity. Triangle symbols below the 1.0 line The rest of this section will expand on some of the im- show fields that have improved relative to HadGEM2-A and provements and degradations highlighted in Fig. 2. Fig- symbols above this line show fields that are degraded. This ure 3 illustrates a comparison of the screen-level (1.5 m) figure shows that only 6 fields, dominated by tropical pre- temperature from GA3.0/GL3.0 meaned over the season cipitation over land, are significantly worse than HadGEM2- June–August (JJA) with that from HadGEM2-A and the A (shown in red); 16 fields, however, are significantly bet- CRUTEM3 observation dataset (Brohan et al., 2006). This ter. The remaining fields either lie within model uncertainty and all following 4-up plots follow a standard layout in which (grey shading) or both models lie within observational uncer- panel (a) shows the mean field from the test, (b) the differ- tainty (whisker bars) where a clear improvement/detriment is ence between the mean test and control, (c) the mean error hard to detect. Of these improved or neutral fields, 14 remain in the control data and (d) the mean error in the test. Fig- ure 3c shows that HadGEM2-A is excessively hot in North

Geosci. Model Dev., 4, 919–941, 2011 www.geosci-model-dev.net/4/919/2011/ D. N. Walters et al.: MetUM GA3.0/3.1 and JULES GL3.0/3.1 configurations 931

a) GA3.0/GL3.0 b) GA3.0/GL3.0 - HadGEM2-A Mean temperature at 1.5m (JJA) Mean temperature at 1.5m (JJA) 90N 90N

60N 60N

30N 30N

0 0

30S 30S

60S 60S

90S 90S 180 150W120W 90W 60W 30W 0 30E 60E 90E 120E 150E 180 150W120W 90W 60W 30W 0 30E 60E 90E 120E 150E Mean = 289.15 K Std dev = 14.33 K Mean diff = -0.31 K RMS diff = 1.08 K

215 230 245 260 275 290 305 320 -12 -9 -6 -3 0 3 6 9 12

c) HadGEM2-A - CRUTEM3 (1979-98) d) GA3.0/GL3.0 - CRUTEM3 (1979-98) Mean temperature at 1.5m (JJA) Mean temperature at 1.5m (JJA) 90N 90N

60N 60N

30N 30N

0 0

30S 30S

60S 60S

90S 90S 180 150W120W 90W 60W 30W 0 30E 60E 90E 120E 150E 180 180 150W120W 90W 60W 30W 0 30E 60E 90E 120E 150E 180 Mean error = 0.95 K RMS error = 2.06 K Mean error = 0.43 K RMS error = 1.55 K

-12 -9 -6 -3 0 3 6 9 12 -12 -9 -6 -3 0 3 6 9 12

Fig. 3.Fig.Climatological 3. Climatological 1.5 m temperatures 1.5 m duringtemperatures JJA in a 30 during yr N96 atmosphere/land-onlyJJA in a 30year N96 simulation.atmosphere/land-only(a) full field from GA3.0/GL3.0, simula- (b) modeltion. difference(a) (GA3.0/GL3.0full field from minus GA3.0/GL3.0, HadGEM2-A), (c)(b)HadGEM2-Amodel difference bias (HadGEM2-A (GA3.0/GL3.0 minus observations) minus HadGEM2-A), and (d) GA3.0/GL3.0(c) bias (GA3.0/GL3.0HadGEM2-A minus observations). bias (HadGEM2-A Observations minus used are observations) from CRUTEM3 and (Brohan(d) GA3.0/GL3.0 et al., 2006). bias (GA3.0/GL3.0 minus

observations). Observationsa) GA3.0/GL3.0 used are from CRUTEM3 (Brohan etb) al.,GA3.0/GL3.0 2006). - HadGEM2-A Mean temperature at 1.5m (DJF) Mean temperature at 1.5m (DJF) 90N 90N

60N 60N

30N 30N

0 50 0

30S 30S

60S 60S

90S 90S 180 150W120W 90W 60W 30W 0 30E 60E 90E 120E 150E 180 150W120W 90W 60W 30W 0 30E 60E 90E 120E 150E Mean = 285.41 K Std dev = 16.55 K Mean diff = -0.05 K RMS diff = 1.18 K

215 230 245 260 275 290 305 320 -12 -9 -6 -3 0 3 6 9 12

c) HadGEM2-A - CRUTEM3 (1979-98) d) GA3.0/GL3.0 - CRUTEM3 (1979-98) Mean temperature at 1.5m (DJF) Mean temperature at 1.5m (DJF) 90N 90N

60N 60N

30N 30N

0 0

30S 30S

60S 60S

90S 90S 180 150W120W 90W 60W 30W 0 30E 60E 90E 120E 150E 180 180 150W120W 90W 60W 30W 0 30E 60E 90E 120E 150E 180 Mean error = -0.53 K RMS error = 2.70 K Mean error = 0.02 K RMS error = 2.21 K

-12 -9 -6 -3 0 3 6 9 12 -12 -9 -6 -3 0 3 6 9 12

Fig. 4. Fig.Climatological 4. Climatological 1.5 m temperatures 1.5 m duringtemperatures DJF in the 30during yr N96 DJF simulation in the using 30year the sameN96 formatsimulation as Fig. 3. using the same format as Fig. 3. www.geosci-model-dev.net/4/919/2011/ Geosci. Model Dev., 4, 919–941, 2011

51 932 D. N. Walters et al.: MetUM GA3.0/3.1 and JULES GL3.0/3.1 configurations

a) GA3.0/GL3.0 b) GA3.0/GL3.0 - HadGEM2-A Mean precipitation rate (JJA) Mean precipitation rate (JJA) 90N 90N

60N 60N

30N 30N

0 0

30S 30S

60S 60S

90S 90S 180 150W120W 90W 60W 30W 0 30E 60E 90E 120E 150E 180 150W120W 90W 60W 30W 0 30E 60E 90E 120E 150E Mean = 3.14 mm/hour Std dev = 3.26 mm/hour Mean diff = -0.03 mm/day RMS diff = 1.21 mm/day

0.1 0.25 0.5 1 1.5 2 2.5 5 7.5 10 15 20 30 40 -20 -15 -10 -5 -2 -1 -0.5 0 0.5 1 2 5 10 15 20

c) HadGEM2-A - GPCP (1979-98) d) GA3.0/GL3.0 - GPCP (1979-98) Mean precipitation rate (JJA) Mean precipitation rate (JJA) 90N 90N

60N 60N

30N 30N

0 0

30S 30S

60S 60S

90S 90S 180 150W120W 90W 60W 30W 0 30E 60E 90E 120E 150E 180 180 150W120W 90W 60W 30W 0 30E 60E 90E 120E 150E 180 Mean error = 0.55 mm/day RMS error = 2.02 mm/day Mean error = 0.52 mm/day RMS error = 1.87 mm/day

-20 -15 -10 -5 -2 -1 -0.5 0 0.5 1 2 5 10 15 20 -20 -15 -10 -5 -2 -1 -0.5 0 0.5 1 2 5 10 15 20

Fig. 5.Fig.Climatological 5. Climatological precipitation precipitation during JJA in during the 30 yr JJA N96 in simulation the 30 year usingN96 the samesimulation format as using Fig. 3. the Observations same format used as are from GPCPFig. (Adler 3. et Observations al., 2003). used are from GPCP (Adler et al., 2003).

America, Europe and Northern Asia. Figure 3b shows that model biases. This improvement is partly due to improve- the change in 1.5 m temperature between HadGEM2-A and ments to soil thermal conductivity in the model (Dharssi GA3.0/GL3.0 has a spatial pattern that is approximately anti- et al., 2009). In GA3.0/GL3.0 the soil is more able to absorb correlated with the HadGEM2-A bias. This means that 52 heat within deep soil layers during the summer and gradually GA3.0/GL3.0 reduces these temperature biases leading to release that heat throughout the winter, warming the surface. a new distribution of bias (Fig. 3d) in which the mean errors Figure 5 shows model precipitation compared with Global over Europe and northern Asia are close to zero and the bias Precipitation Climatology Project (GPCP) data (Adler et al., over North America is significantly reduced. As each change 2003) during JJA. In HadGEM2-A there are a range of model to the model configuration since HadGEM2-A has been as- precipitation biases, including wet biases over the equato- sessed in a 10 yr N96 AMIP run, it is possible to suggest rial Indian Ocean and the Western Pacific, and dry biases an attribution of some changes in performance to particular over India, the Maritime Continent (Indonesia and surround- model enhancements. The physics change that appears to ing islands) and the Sahel region of Africa. Most of these contribute most to these improvements in JJA temperatures errors have reduced in GA3.0/GL3.0 with the exception of is the use of van Genuchten soil hydraulics (van Genuchten, the dry bias over India. Despite the under-prediction of 1980) instead of those of Clapp and Hornberger (1978) and rain over India, we have seen an improvement in the In- the replacement of soil properties derived from the Interna- dian monsoon winds in GA3.0/GL3.0 (not shown). The dry tional Geosphere-Biosphere Programme’s Global Soil Data bias over the Maritime Continent has been reduced mostly Task (Global Soil Data Task, 2000) with those derived from due to the inclusion of a “buddy” scheme for coastal grid- the Harmonised World Soil Database (Nachtergaele et al., points which uses an average wind speed over neighbour- 2008). This change allows the soil to retain more moisture ing sea points to split the level 1 wind speed into separate near the surface, leading to increased latent heating and de- land and sea contributions. This enhances the wind speed creased surface temperatures. Figure 4 shows the equiva- over the sea part of the grid-box giving improved scalar lent plot for 1.5 m temperatures during December to Febru- fluxes there. The wet bias in the Western Pacific has also re- ary (DJF). In this season HadGEM2-A is excessively cold duced in GA3.0/GL3.0. This is due to a range of convection over North America and Central Asia. The GA3.0/GL3.0 parametrization changes including reducing the amount of simulation is warmer in these locations leading to reduced convective momentum transport (Stratton et al., 2009) and

Geosci. Model Dev., 4, 919–941, 2011 www.geosci-model-dev.net/4/919/2011/ D. N. Walters et al.: MetUM GA3.0/3.1 and JULES GL3.0/3.1 configurations 933 increasing the timescale over which CAPE is removed by 6.2 Enhanced resolution AMIP climate simulations the convection scheme. The equatorial Indian Ocean and the Sahel region of Africa have also seen their biases reduce due In addition to the 30 yr N96 simulation discussed above, to the use of the PC2 cloud scheme, which generates a net Fig. 2 shows the normalised assessment of a similar 30 yr reduction in clouds and precipitation over the tropical oceans simulation at N216 resolution (approximately 60 km in the and a net increase in clouds and precipitation over tropical mid-latitudes) and a 20 yr simulation (1979 to 1998) at N320 land. This increases precipitation over the Sahel (reducing resolution (approximately 40 km in the mid-latitudes). This the bias) and decreases precipitation over the equatorial In- shows that in general, there are no widespread changes to dian Ocean (also reducing the bias). Figure 6 shows the the basic atmospheric state in response to the increased hor- equivalent plot for precipitation during DJF. Many of the im- izontal resolution when assessed against these low resolu- provements seen in JJA are also seen in DJF including the tion climatologies. This is a reassuring confirmation that increased precipitation over the Maritime Continent and the the performance of the GA3.0/GL3.0 configuration is trace- reduced precipitation over the western Pacific and equato- able across this range of resolutions, which is a principle rial Indian Ocean. One improvement that is specific to DJF on which the MetUM Global Atmosphere development pro- occurs in the North Atlantic to the west of the UK where cess relies. The increased resolution of these simulations, HadGEM2-A has a wet bias. GA3.0/GL3.0 has reduced the however, does enable the model to represent both regional amount of precipitation in this region to such an extent that and local climatology better, e.g. by improving precipitation there is now a slight dry bias. There have been many physics distributions and amounts, especially near areas of orogra- changes introduced since HadGEM2-A that have affected the phy. Enhanced resolution allows the more accurate mod- model’s performance in the North Atlantic, which makes it elling of extremes, including the number, intensity, and in- difficult to attribute the cause of this change. terannual variability of tropical cyclones, particularly in the As stated earlier, many top-of-atmosphere radiation fields Atlantic. Atmosphere/land-only simulations in other mod- have improved between HadGEM2-A and GA3.0/GL3.0. els have also shown stronger large scale motion and asso- Figure 7 shows the improvements in outgoing shortwave ra- ciated precipitation on the southern side of the Gulf Stream diation during the DJF season (i.e. the amount of the Sun’s due to the atmosphere being able to react to strong SST gra- radiation reflected out to space by clouds and the Earth’s sur- dients (Minobe et al., 2008). In previous simulations of cou- face) compared to the Clouds and the Earth’s Radiant Energy pled modelling systems, enhanced horizontal atmospheric System (CERES) dataset (Wielicki et al., 1996). HadGEM2- resolution has been shown to lead to an improved El Nino-˜ A reflects too much of the Sun’s radiation in two small re- Southern Oscillation (Shaffrey et al., 2009); it has also been gions located to the west of Peru and to the west of Southern seen to contribute to improvements in simulations of Atlantic Africa; this is because it generates too much stratocumulus blocking, again by being able to better respond to improved cloud in these regions and the thickness of this stratocumulus SSTs (Scaife et al., 2011). Figure 2 shows that at N216 and cloud is too uniform throughout the grid-box. These outgo- N320 we see an increase in the precipitation errors over trop- ing shortwave biases have been reduced in GA3.0/GL3.0 due ical land compared to N96, although we have also seen an to the following physics changes: (i) the PC2 cloud scheme, improvement to the rain deficit over India during JJA (not which represents different cloud regimes´ as a different bal- shown). There are also improvements in the precipitable wa- ance between cloud creation and destruction processes; (ii) ter and relative humidity diagnostics. Looking at this in more enhancing the turbulent entrainment mixing in stratocumulus detail, it can be seen in DJF and JJA zonal mean relative hu- over cumulus, which reduces the amount of stratocumulus midity plots (not shown) that the higher resolution models cloud and outgoing shortwave and (iii) the McICA cloud in- tend to increase the relative humidity through most of the homogeneity, which allows more shortwave radiation to pass troposphere in the tropics, and decrease it at mid to high lati- through the thinner parts of the cloud, decreasing the amount tudes, both of which are generally an improvement. This ap- of shortwave reflected back to space (Hill et al., 2011). We pears to be a robust result of increased resolution, as it is seen can see, however, a detrimental reduction in reflected solar at both N320 and N216, and was also seen in comparing dif- radiation along the Intertropical Convergence Zone in the Pa- ferent resolution models during previous model assessments. cific. Figure 7 also shows a negative bias in the Southern The difference is most likely due to increased intensity of Ocean off the coast of Antarctica. This is associated with tropical deep convection in the higher resolution model, per- increased surface shortwave and leads to a Southern Ocean haps associated with a slight increase in the strength of the warming when GA3.0/GL3.0 are coupled to an ocean model Hadley Circulation. (not shown). This bias is slightly reduced in GA3.0/GL3.0 due to prognostic rain with Abel and Shipway (2007) fall 6.3 Short-range NWP case studies speeds allowing the slower droplets to evaporate more ef- ficiently, which increases relative humidity and hence the Finally, we present some initial results from a comparison cloud amount. of GA3.0/GL3.0 and GA3.1/GL3.1 in a set of 20 NWP case studies made up of 10 cases in the season December– www.geosci-model-dev.net/4/919/2011/ Geosci. Model Dev., 4, 919–941, 2011 934 D. N. Walters et al.: MetUM GA3.0/3.1 and JULES GL3.0/3.1 configurations

a) GA3.0/GL3.0 b) GA3.0/GL3.0 - HadGEM2-A Mean precipitation rate (DJF) Mean precipitation rate (DJF) 90N 90N

60N 60N

30N 30N

0 0

30S 30S

60S 60S

90S 90S 180 150W120W 90W 60W 30W 0 30E 60E 90E 120E 150E 180 150W120W 90W 60W 30W 0 30E 60E 90E 120E 150E Mean = 3.02 mm/hour Std dev = 2.95 mm/hour Mean diff = -0.03 mm/day RMS diff = 0.89 mm/day

0.1 0.25 0.5 1 1.5 2 2.5 5 7.5 10 15 20 30 40 -20 -15 -10 -5 -2 -1 -0.5 0 0.5 1 2 5 10 15 20

c) HadGEM2-A - GPCP (1979-98) d) GA3.0/GL3.0 - GPCP (1979-98) Mean precipitation rate (DJF) Mean precipitation rate (DJF) 90N 90N

60N 60N

30N 30N

0 0

30S 30S

60S 60S

90S 90S 180 150W120W 90W 60W 30W 0 30E 60E 90E 120E 150E 180 180 150W120W 90W 60W 30W 0 30E 60E 90E 120E 150E 180 Mean error = 0.44 mm/day RMS error = 1.54 mm/day Mean error = 0.41 mm/day RMS error = 1.43 mm/day

-20 -15 -10 -5 -2 -1 -0.5 0 0.5 1 2 5 10 15 20 -20 -15 -10 -5 -2 -1 -0.5 0 0.5 1 2 5 10 15 20

Fig. 6.Fig.Climatological 6. Climatological precipitation precipitation during DJF in the during 30 yr DJFN96 simulation in the 30year using theN96 samesimulation format as Fig. using3. the same format as Fig. 3.

a) GA3.0/GL3.0 b) GA3.0/GL3.0 - HadGEM2-A Mean outgoing sw rad flux (toa) (DJF) Mean outgoing sw rad flux (toa) (DJF) 90N 90N

60N 60N

30N 30N 0 53 0 30S 30S

60S 60S

90S 90S 180 150W120W 90W 60W 30W 0 30E 60E 90E 120E 150E 180 150W120W 90W 60W 30W 0 30E 60E 90E 120E 150E Mean = 103.50 W/m2 Std dev = 54.29 W/m2 Mean diff = 0.26 W/m2 RMS diff = 8.35 W/m2

0 60 120 180 240 300 -45 -30 -15 0 15 30 45

c) HadGEM2-A - CERES SRBAVG d) GA3.0/GL3.0 - CERES SRBAVG Mean outgoing sw rad flux (toa) (DJF) Mean outgoing sw rad flux (toa) (DJF) 90N 90N

60N 60N

30N 30N

0 0

30S 30S

60S 60S

90S 90S 180 150W120W 90W 60W 30W 0 30E 60E 90E 120E 150E 180 150W120W 90W 60W 30W 0 30E 60E 90E 120E 150E Mean error = 0.59 W/m2 RMS error = 15.68 W/m2 Mean error = 0.85 W/m2 RMS error = 15.04 W/m2

-45 -30 -15 0 15 30 45 -45 -30 -15 0 15 30 45

Fig. 7.Fig.Climatological 7. Climatological outgoing shortwave outgoing radiation shortwave from the radiation top of the fromatmosphere the top during of DJF thein atmosphere the 30 yr N96 during simulation DJF using in the the same format30year as Fig. 3.N96 Observationssimulation used usingare from the CERES same (Wielicki format et as al. Fig., 1996 3.). Observations used are from CERES (Wielicki et al., 1996).

Geosci. Model Dev., 4, 919–941, 2011 www.geosci-model-dev.net/4/919/2011/

54 D. N. Walters et al.: MetUM GA3.0/3.1 and JULES GL3.0/3.1 configurations 935

a) GA3.1/GL3.1 forecast b) GA3.1/GL3.1 forecast - GA3.0/GL3.0 forecast T+24 temperature at 1.5m (JJA) T+24 temperature at 1.5m (JJA) 90N 90N

60N 60N

30N 30N

0 0

30S 30S

60S 60S

90S 90S 180 150W120W 90W 60W 30W 0 30E 60E 90E 120E 150E 180 150W120W 90W 60W 30W 0 30E 60E 90E 120E 150E Mean = 289.75 K Std dev = 14.63 K Mean diff = 0.07 K RMS diff = 0.41 K

215 230 245 260 275 290 305 320 -2.25 -1.5 -0.75 0 0.75 1.5 2.25

c) GA3.0/GL3.0 forecast - Operational analysis d) GA3.1/GL3.1 forecast - Operational analysis T+24 temperature at 1.5m (JJA) T+24 temperature at 1.5m (JJA) 90N 90N

60N 60N

30N 30N

0 0

30S 30S

60S 60S

90S 90S 180 150W120W 90W 60W 30W 0 30E 60E 90E 120E 150E 180 150W120W 90W 60W 30W 0 30E 60E 90E 120E 150E Mean error = -0.26 K RMS error = 0.81 K Mean error = -0.19 K RMS error = 0.62 K

-2.25 -1.5 -0.75 0 0.75 1.5 2.25 -2.25 -1.5 -0.75 0 0.75 1.5 2.25

Fig. 8.Fig.Mean 8. Mean 24 h 24hourforecast 1.5forecast m temperatures 1.5 m temperatures during 10 N320 during NWP 10 N320 case studiesNWP in case JJA. studies(a) full in fieldJJA. from(a) full GA3.1/GL3.1, field (b) modelfrom difference GA3.1/GL3.1, (GA3.1/GL3.1(b) model minus GA3.0/GL3.0), difference (GA3.1/GL3.1(c) GA3.0/GL3.0 minus bias (GA3.0/GL3.0 GA3.0/GL3.0), minus(c) MetGA3.0/GL3.0 Office operational bias analyses) and (d)(GA3.0/GL3.0GA3.1/GL3.1 bias minus (GA3.1/GL3.1 Met Office minus operational analyses). analyses) and (d) GA3.1/GL3.1 bias (GA3.1/GL3.1 minus analyses). February (DJF) spread across DJF 2008/9 and 2009/10 and and 00:00 UTC (T+12, 18, 36, 42 etc.) have a warm bias 10 cases in June–August (JJA) spread across JJA 2008 and of about 0.5 ◦C in GA3.1/GL3.1. This is exacerbated in 2009. The cases are initialised using Met Office operational GA3.0/GL3.0, which as with 12:00 UTC is between 0.2 and global NWP analyses valid at 12:00 UTC and are chosen so 55 0.5 ◦C warmer than GA3.1/GL3.1. Figure 9b shows verifi- that subsequent start dates are separated by 14 days to min- cation of the T+24 temperature profile against radiosondes imise the synoptic correlation between cases. Each case is released from Northern Hemisphere land stations. As ex- run at N320 resolution for a period of 5 days. pected, the main difference is constrained to the lower tro- Figure 8 illustrates a comparison between the mean posphere, since the vast majority of differences between the 24 h forecast (T + 24) screen-level temperature from the GA3.0/GL3.0 and GA3.1/GL3.1 configurations described in 10 GA3.0/GL3.0 and GA3.1/GL3.1 cases during JJA. Sect. 5 are in the land surface model and the boundary layer GA3.1/GL3.1 is significantly warmer than GA3.0/GL3.0 scheme. As in Fig. 9a, we see that GA3.0/GL3.0 is warmer over Antarctica, due to the longer tail in the stability func- near the surface and the biases at 925 hPa are roughly con- tion used for stable boundary layers. In the Northern Hemi- sistent with the screen-level results. At 1000 hPa, how- sphere, there is a cooling over land in Europe and North ever, biases are either side of zero with the magnitude of America, which we can see reduces the warm bias present in the bias in GA3.1/GL3.1 being only marginally smaller than GA3.0/GL3.0. Sensitivity tests have shown that the land sur- that in GA3.0/GL3.0. Figure 10 shows the T+24 screen- face is systematically cooler with the single amalgamated tile level temperature from GA3.0/GL3.0 and GA3.1/GL3.1 dur- rather than 9 separate tiles, whilst the other land and bound- ing DJF. As with the JJA results in Fig. 8, we see a signif- ary layer changes contribute to the total signal in a non-trivial icant warming near the winter pole and an average cooling way. Figure 9a shows the verification of these fields against over land in the summer hemisphere. Other near-surface land-based surface observations in the Northern Hemisphere forecast fields from GA3.1/GL3.1 have correlated improve- as a function of forecast range. This confirms that after ments when compared with GA3.0/GL3.0, such as a de- 24 h, the GA3.1/GL3.1 bias is very close to zero, whilst the creased negative bias in pressure at mean sea level over GA3.0/GL3.0 bias is between 0.1–0.2 ◦C and is up to 0.3 ◦C Northern Hemisphere land in JJA (not shown). It is for this after 5 days. Furthermore, the forecasts valid at 18:00 UTC reason that GA3.1/GL3.1 rather than GA3.0/GL3.0 has been www.geosci-model-dev.net/4/919/2011/ Geosci. Model Dev., 4, 919–941, 2011 936 D. N. Walters et al.: MetUM GA3.0/3.1 and JULES GL3.0/3.1 configurations

-200 a) 1.5 b) 0.0 GA/GL3.0

0 GA/GL3.1 200.0 1.0

200

400.0 400

0.5

600 600.0 Pressure (hPa) FC-Obs Mean Error (K)

800

0.0 -50 0 50 100 150 800.0

1000

-0.5 1000.0 1200 -12 0 12 24 36 48 60 72 84 96108120132 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 Forecast Range (hours) FC-Obs Mean Error (K)

Fig.Fig. 9. Mean 9. Mean forecast forecast error for 1.5 error m temperature for 1.5 asm a functiontemperature of forecast as range a function(a) and radiosonde of forecast temperature range profiles(a) and after radiosonde a 24 h forecast (b)temperaturefrom the 10 N320 profiles NWPafter case studies a 24 in h JJA. forecast These errors(b) from are meaned the 10 overN320 all land-basedNWP case observations studies in thein JJA. extra-tropical These Northern errors Hemisphere (20◦–90◦ N). are meaned over all land-based observations in the extra-tropical Northern Hemisphere (20◦–90◦ N).

a) GA3.1/GL3.1 forecast b) GA3.1/GL3.1 forecast - GA3.0/GL3.0 forecast T+24 temperature at 1.5m (DJF) T+24 temperature at 1.5m (DJF) 90N 90N

60N 60N

30N 30N

0 0

30S 30S

60S 5660S

90S 90S 180 150W120W 90W 60W 30W 0 30E 60E 90E 120E 150E 180 150W120W 90W 60W 30W 0 30E 60E 90E 120E 150E Mean = 286.35 K Std dev = 16.06 K Mean diff = 0.08 K RMS diff = 0.43 K

215 230 245 260 275 290 305 320 -2.25 -1.5 -0.75 0 0.75 1.5 2.25

c) GA3.0/GL3.0 forecast - Operational analysis d) GA3.1/GL3.1 forecast - Operational analysis T+24 temperature at 1.5m (DJF) T+24 temperature at 1.5m (DJF) 90N 90N

60N 60N

30N 30N

0 0

30S 30S

60S 60S

90S 90S 180 150W120W 90W 60W 30W 0 30E 60E 90E 120E 150E 180 150W120W 90W 60W 30W 0 30E 60E 90E 120E 150E Mean error = -0.16 K RMS error = 0.68 K Mean error = -0.09 K RMS error = 0.51 K

-2.25 -1.5 -0.75 0 0.75 1.5 2.25 -2.25 -1.5 -0.75 0 0.75 1.5 2.25

Fig. 10.Fig.Mean 10. 24Mean h forecast 24hour 1.5 m temperatures forecast 1.5 fromm thetemperatures 10 N320 NWP from case studies the 10 inN320 DJF usingNWP the samecase format studies as Fig.in DJF8. using the same format as Fig. 8.

Geosci. Model Dev., 4, 919–941, 2011 www.geosci-model-dev.net/4/919/2011/

57 D. N. Walters et al.: MetUM GA3.0/3.1 and JULES GL3.0/3.1 configurations 937 implemented in the deterministic global forecast component (approximately 135 km) to N512 (25 km) and above, we aim of the Met Office operational NWP suite. It is worth noting, to provide a trusted configuration that can be used for re- however, that the initial analyses used in these case studies search within the Met Office, UK academia and the growing were generated by a system using surface and boundary layer community of MetUM collaborators around the world. parametrizations much closer to those in GA3.1/GL3.1 than GA3.0/GL3.0. There is a possibility that this is related to the Acknowledgements. MB, DC, PF, CH, GM, WM-O, MP, MR, relative good performance of GA3.1/GL3.1 when initialised AS and KW were supported by the Joint DECC/Defra Met Office Hadley Centre Climate Programme (GA01101). The development from these analyses, and in particular the soil moisture anal- of the Global Atmosphere/Land configurations was possible only yses that at the time used only screen-level observations of through the hard work of a large number of people, both within and temperature and humidity to infer errors in the soil mois- outside the Met Office, that exceeds the list of authors. ture (Best and Maisey, 2002). We are currently performing Intellectual property. full NWP system trials of GA3.0/GL3.0 and GA3.1/GL3.1, Due to intellectual property right restrictions, we cannot provide including lower resolution trials run over several months, either the source code or documentation papers for the MetUM or which will allow the model’s analysis cycle to respond to the JULES. changes in model characteristics. This will help us to assess Obtaining the MetUM. the true relative benefits of GA3.0/GL3.0 and GA3.1/GL3.1 The Met Office Unified Model is available for use under licence. in NWP systems and provide insight into how best to resolve A number of research organisations and national meteorological the differences between the two configurations in the coming services use the MetUM in collaboration with the Met Office to undertake basic atmospheric process research, produce forecasts, year. develop the MetUM code and build and evaluate Earth System mod- els. For further information on how to apply for a licence see http: //www.metoffice.gov.uk/research/collaboration/um-collaboration 7 Summary and conclusions Obtaining JULES. JULES is available under licence free of charge. Please The primary purpose of this paper is to provide a detailed de- contact a JULES partner institution for permission to use scription of the scientific formulation of Global Atmosphere JULES for research purposes. For a list of JULES partners see 3.0 and Global Land 3.0 (GA3.0/GL3.0) and the related http://www.jchmr.org/jules/licence.html Global Atmosphere 3.1 and Global Land 3.1 (GA3.1/GL3.1) Copyright statement. configurations, for those utilising or contributing to them This work is distributed under the Creative Commons Attribution throughout the MetUM and JULES communities. We have 3.0 License together with an author copyright. This license does presented results from some initial assessments showing en- not conflict with the regulations of the Crown Copyright. couraging performance of GA3.0/GL3.0 in atmosphere/land- Edited by: J. C. Hargreaves only climate integrations, but have shown that GA3.1/GL3.1 performs better in NWP case studies initialised from Met Office operational analyses. We have listed the primary ap- References plications in which the Met Office have either implemented GA3.0/GL3.0 or GA3.1/GL3.1 or plans to do so. Abel, S. J. and Shipway, B. J.: A comparison of cloud-resolving We have also discussed the potential benefits of a unified model simulations of trade wind cumulus with aircraft observa- and open development process. By unifying the Global At- tions taken during RICO, Q. J. Roy. Meteorol. Soc., 133, 781– 794, 2007. mosphere, future enhancements will only be included if they Abel, S. J., Walters, D. N., and Allen, G.: Evaluation of stra- prove acceptable across all timescales. The resulting config- tocumulus cloud prediction in the Met Office forecast model uration will, therefore, have a higher standard of scientific during VOCALS-REx, Atmos. Chem. Phys., 10, 10541–10559, integrity due to being verified against real-time observations, doi:10.5194/acp-10-10541-2010, 2010. atmospheric data assimilation analyses and high quality long Adler, R. F., Huffman, G. J., Chang, A., Ferraro, R., Xie, P.-P., time-series climatology datasets. Also, using this combined Janowiak, J., Rudolf, B., Schneider, U., Curtis, S., Bolvin, D., framework to investigate model problems or deficiencies will Gruber, A., Susskind, J., Arkin, P., and Nelkin, E.: The version-2 provide a better understanding of the underlying processes Global Precipitation Climatology Project (GPCP) monthly pre- and allow us to develop robust and scientifically sound so- cipitation analysis (1979-present), J. Hydrometeor., 4, 1147– lutions. Adopting the configuration throughout the Met Of- 1167, 2003. fice’s operational forecasting and climate research activities Allan, R. and Ansell, T.: A new globally complete monthly histor- ical gridded mean sea level pressure dataset (HadSLP2): 1850– will contribute to one of its wider aims of constructing a truly 2004, J. Climate, 19, 5816–5842, 2006. seamless operational weather and climate service. Finally, Arakawa, A. and Lamb, V. R.: Computational design of the ba- by assessing and documenting the performance of the con- sic dynamic processes of the UCLA general circulation model, figuration coupled to other Earth System component mod- Methods Comput. Phys., 17, 173–265, 1977. els and data assimilation systems, from timescales of days Arribas, A., Glover, M., Maidens, A., Peterson, K., Gordon, M., to decades ahead and across horizontal resolutions from N96 MacLachlan, C., Graham, R., Fereday, D., Camp, J., Scaife, www.geosci-model-dev.net/4/919/2011/ Geosci. Model Dev., 4, 919–941, 2011 938 D. N. Walters et al.: MetUM GA3.0/3.1 and JULES GL3.0/3.1 configurations

A. A., Xavier, P., McLean, P., Colman, A., and Cusack, S.: The scheme in the Met Office numerical weather prediction model, GloSea4 ensemble prediction system for seasonal forecasting, Bound.-Lay. Meteorol., 128, 117–132, 2008. Mon. Weather Rev., 139, 1891–1910, 2011. Bushell, A. C., Wilson, D. R., and Gregory, D.: A description of Arribas, A. et al.: A fully resolved stratosphere and sea-ice initiali- cloud production by non-uniformly distributed processes, Q. J. sation: upgrades to the GloSea4 seasonal forecasting system, in Roy. Meteorol. Soc., 129, 1435–1455, 2003. preparation, 2011. Charney, J. G. and Phillips, N. A.: Numerical integration of the Balsamo, G., Pappenberger, F., Dutra, E., Viterbo, P., and van den quasi-geostrophic equations for barotropic and simple baroclinic Hurk, B.: A revised land hydrology in the ECMWF model: a flows, J. Meteor., 10, 71–99, 1953. step towards daily water flux prediction in a fully-closed water Clapp, R. B. and Hornberger, G. M.: Empirical equations for some cycle, Hydrol. Process., 25, 1046–1054, 2011. soil hydraulic properties, Water Resour. Res., 14, 601–604, 1978. Barker, H. W. and Li, Z.: Improved simulation of clear-sky short- Clark, D. B., Mercado, L. M., Sitch, S., Jones, C. D., Gedney, N., wave radiative transfer in the CCC-GCM, J. Climate, 8, 2213– Best, M. J., Pryor, M., Rooney, G. G., Essery, R. L. H., Blyth, 2223, 1995. E., Boucher, O., Harding, R. J., Huntingford, C., and Cox, P. Beard, K. V.: Terminal velocity and shape of cloud and precipitation M.: The Joint UK Land Environment Simulator (JULES), model drops aloft, J. Atmos. Sci., 33, 851–864, 1976. description – Part 2: Carbon fluxes and vegetation dynamics, Beljaars, A. C. M. and Holtslag, A. A. M.: Flux parametrization Geosci. Model Dev., 4, 701–722, doi:10.5194/gmd-4-701-2011, over land surfaces for atmospheric models, J. Appl. Meteorol., 2011. 30, 327–341, 1991. Clough, S. A., Kneizys, F. X., and Davies, R. W.: Line shape and Beljaars, A. C. M. and Viterbo, P.: Role of the boundary layer in the water vapor continuum, Atmos. Res., 23, 229–241, 1989. a numerical weather prediction model., in: Clear and cloudy Coleman, K. and Jenkinson, D. S.: RothC-26.3, a model for the boundary layers, edited by: Holtslag, A. A. M. and Duynkerke, turnover of carbon in soil: Model description and user’s guide, P. G., Royal Netherlands Academy of Arts and Sciences, Ams- Tech. rep., Lawes Agricultural Trust, IACR Rothamsted, Harp- terdam, 1998. enden, UK, 1999. Berrisford, P., Dee, D., Fielding, K., Fuentes, M., Kallberg,˚ P., Collins, W. J., Bellouin, N., Doutriaux-Boucher, M., Gedney, N., Kobayashi, S., and Uppala, S.: The ERA-Interim archive, Tech. Hinton, T., Jones, C. D., Liddicoat, S., Martin, G., O’Connor, Rep. 1, ERA report series, ECMWF, Reading, UK, 2009. F., Rae, J., Senior, C., Totterdell, I., Woodward, S., Reichler, T., Best, M. J.: Representing urban areas within operational numerical Kim, J., and Halloran, P.: Evaluation of the HadGEM2 model, weather prediction models, Bound.-Lay. Meteorol., 114, 91–109, Tech. Rep. 74, Hadley Centre, Met Office, Exeter, UK, 2008. 2005. Cox, P.: Description of the TRIFFID dynamic global vegetation Best, M. J. and Maisey, P. E.: A physically based soil moisture model, Tech. Rep. 24, Hadley Centre, Met Office, Bracknell, nudging scheme, Tech. Rep. 35, Hadley Centre, Met Office, UK, 2001. Bracknell, UK, 2002. Cox, P., Betts, R., Jones, C., Spall, S., and Totterdell, I.: Accel- Best, M. J., Pryor, M., Clark, D. B., Rooney, G. G., Essery, R .L. eration of global warming due to carbon-cycle feedbacks in a H., Menard,´ C. B., Edwards, J. M., Hendry, M. A., Porson, A., coupled climate model, Nature, 408, 184–187, 2000. Gedney, N., Mercado, L. M., Sitch, S., Blyth, E., Boucher, O., Cox, P. M., Betts, R. A., Bunton, C. B., Essery, R., Rowntree, P. R., Cox, P. M., Grimmond, C. S. B., and Harding, R. J.: The Joint , and Smith, J.: The impact of new land surface physics on the UK Land Environment Simulator (JULES), model description – GCM simulation of climate and climate sensitivity, Clim. Dy- Part 1: Energy and water fluxes, Geosci. Model Dev., 4, 677– nam., 15, 183–203, 1999. 699, doi:10.5194/gmd-4-677-2011, 2011. Cullen, M. J. P.: The unified forecast/climate model, Meteor. Mag., Beven, K. and Kirkby, M. J.: A physically based, variable contribut- 122, 81–94, 1993. ing area model of basin hydrology, Hydrol. Sci. B., 24, 43–69, Cusack, S., Slingo, A., Edwards, J. M., and Wild, M.: The radia- 1979. tive impact of a simple aerosol climatology on the Hadley Centre Bosilovich, M. G.: NASA0s modern era retrospectiveanalysis atmospheric GCM, Q. J. Roy. Meteorol. Soc., 124, 2517–2526, for research and applications: Integrating Earth obser- 1998. vations, Earthzine, http://www.earthzine.org/2008/09/26/ Cusack, S., Edwards, J. M., and Crowther, J. M.: Investigating k nasas-modern-era-retrospective-analysis, last access: 12 April distribution methods for parameterizing gaseous absorption in 2011, 2008. the Hadley Centre climate model, J. Geophys. Res., 104, 2051– Bowler, N. E., Arribas, A., Mylne, K. R., Robertson, K. B., and 2057, 1999. Beare, S. E.: The MOGREPS short-range ensemble prediction Davies, T., Cullen, M. J. P., Malcolm, A. J., Mawson, M. H., Stan- system, Q. J. Roy. Meteorol. Soc., 134, 703–722, 2008. iforth, A., White, A. A., and Wood, N.: A new dynamical core Brohan, P., Kennedy, J. J., Harris, I., Tett, S. F. B., and Jones, P. D.: for the Met Office’s global and regional modelling of the atmo- Uncertainty estimates in regional and global observed tempera- sphere, Q. J. Roy. Meteorol. Soc., 131, 1759–1782, 2005. ture changes: a new dataset from 1850, J. Geophys. Res., 111, Derbyshire, S. H., Maidens, A. V., Milton, S. F., Stratton, R. A., D12106, doi:10.1029/2005JD006548, 2006. and Willett, M. R.: Adaptive detrainment in a convective Brown, A. R. and Webster, S.: Orographic flow-blocking scheme parametrization., Q. J. Roy. Meteorol. Soc., doi:10.1002/qj.875, characteristics., Q. J. Roy. Meteorol. Soc., 130, 3015–3028, in press, 2011. 2004. Dharssi, I., Vidale, P. L., Verhoef, A., Macpherson, B., Jones, C., Brown, A. R., Beare, R. J., Edwards, J. M., Lock, A. P., Keogh, S. J., and Best, M.: New soil physical properties implemented in the Milton, S. F., and Walters, D. N.: Upgrades to the boundary-layer Unified Model at PS18, Tech. Rep. 528, Forecasting R&D, Met

Geosci. Model Dev., 4, 919–941, 2011 www.geosci-model-dev.net/4/919/2011/ D. N. Walters et al.: MetUM GA3.0/3.1 and JULES GL3.0/3.1 configurations 939

Office, Exeter, UK, 2009. wave breaking: Impact upon the climate of the UK Meteorolog- Donlon, C. J., Martin, M., Stark, J. D., Roberts-Jones, J., and ical Office Unified Model, Q. J. Roy. Meteorol. Soc., 124, 463– Fiedler, E.: The Operational Sea Surface Temperature and 493, 1998. Sea Ice Analysis (OSTIA) system, Remote Sens. Environ., HadGEM2 Development Team: Martin, G. M., Bellouin, N., doi:10.1016/j.rse.2010.10.017, in press, 2011. Collins, W. J., Culverwell, I. D., Halloran, P. R., Hardiman, S. C., Dyer, A. J. and Hicks, B. B.: Flux-gradient relationships in the con- Hinton, T. J., Jones, C. D., McDonald, R. E., McLaren, A. J., stant flux layer, Q. J. Roy. Meteorol. Soc., 96, 715–721, 1970. O’Connor, F. M., Roberts, M. J., Rodriguez, J. M., Woodward, Edwards, J. M.: Efficient calculation of infrared fluxes and cooling S., Best, M. J., Brooks, M. E., Brown, A. R., Butchart, N., Dear- rates using the two-stream equations, J. Atmos. Sci., 53, 1921– den, C., Derbyshire, S. H., Dharssi, I., Doutriaux-Boucher, M., 1932, 1996. Edwards, J. M., Falloon, P. D., Gedney, N., Gray, L. J., Hewitt, Edwards, J. M. and Slingo, A.: Studies with a flexible new radiation H. T., Hobson, M., Huddleston, M. R., Hughes, J., Ineson, S., code. I: Choosing a configuration for a large-scale model, Q. J. Ingram, W. J., James, P. M., Johns, T. C., Johnson, C. E., Jones, Roy. Meteorol. Soc., 122, 689–719, 1996. A., Jones, C. P., Joshi, M. M., Keen, A. B., Liddicoat, S., Lock, Edwards, J. M., Havemann, S., Thelen, J.-C., and Baran, A. J.: A A. P., Maidens, A. V., Manners, J. C., Milton, S. F., Rae, J. G. L., new parametrization for the radiative properties of ice crystals: Ridley, J. K., Sellar, A., Senior, C. A., Totterdell, I. J., Verhoef, Comparison with existing schemes and impact in a GCM, Atmos. A., Vidale, P. L., and Wiltshire, A.: The HadGEM2 family of Res., 83, 19–35, 2007. Met Office Unified Model climate configurations, Geosci. Model Essery, R. L. H., Best, M. J., Betts, R. A., Cox, P. M., and Taylor, Dev., 4, 723–757, doi:10.5194/gmd-4-723-2011, 2011. C. M.: Explicit representation of subgrid heterogeneity in a GCM Haywood, J. M., Allan, R. P., Culverwell, I., Slingo, A., Mil- land surface scheme, J. Hydrometeorol., 4, 530–543, 2003. ton, S., Edwards, J. M., and Clerbaux, N.: Can desert dust Falloon, P. D. and Betts, R. A.: The impact of climate change on explain the outgoing longwave radiation anomaly over the global river flow in HadGEM1 simulations, Atmos. Sci. Lett., 7, Sahara during July 2003?, J. Geophys. Res., 110, D05105, 62–68, 2006. doi:10.1029/2004JD005232, 2005. Fritsch, J. M. and Chappell, C. F.: Numerical prediction of con- Heming, J. T.: The impact of resolution on Met Office model pre- vectively driven mesoscale pressure systems. Part I: convective dictions of tropical cyclone track and intensity, in: AMS 29th parameterization, J. Atmos. Sci., 37, 1722–1733, 1980. Conference on Hurricanes and Tropical Meteorology, Tucson, Gedney, N. and Cox, P. M.: The sensitivity of global climate model Arizona, 10–14th May 2010, 10C.1, 2010. simulations to the representation of soil moisture heterogeneity, Hewitt, H. T., Copsey, D., Culverwell, I. D., Harris, C. M., Hill, R. J. Hydrometeorol., 4, 1265–1275, 2003. S. R., Keen, A. B., McLaren, A. J., and Hunke, E. C.: Design Gedney, N., Cox, P. M., and Huntingford, C.: Climate feed- and implementation of the infrastructure of HadGEM3: the next- back from wetland methane emissions, Geophys. Res. Lett., 31, generation Met Office climate modelling system, Geosci. Model L20503, doi:10.1029/2004GL020919, 2004. Dev., 4, 223–253, doi:10.5194/gmd-4-223-2011, 2011. Global Soil Data Task: Global soil data products CD-ROM Hill, P. G., Manners, J., and Petch, J. C.: Reducing noise associated (IGBP-DIS), CD-ROM, http://www.daac.ornl.gov, International with the Monte Carlo Independent Column Approximation for Geosphere-Biosphere Programme, Data and Information Sys- weather forecasting models, Q. J. Roy. Meteorol. Soc., 137, 219– tem, Potsdam, Germany. Available from Oak Ridge National 228, 2011. Laboratory Distributed Active Archive Center, Oak Ridge, TN, Hogan, R. J. and Illingworth, A. J.: Deriving cloud overlap statistics last access: 12 April 2011, 2000. from radar, Q. J. Roy. Meteorol. Soc., 126, 2903–2909, 2000. Godfrey, J. S. and Beljaars, A. C. M.: On the turbulent fluxes of Hunke, E. C. and Lipscombe, W. H.: CICE: the Los Alamos sea ice buoyancy, heat and moisture at the air-sea interface at low wind model documentation and software user’s manual, Version 4.0, speeds, J. Geophys. Res., 96, 22043–22048, 1991. LA-CC-06-012, Los Alamos National Laboratory, New Mexico, Grant, A. L. M.: Cloud-base fluxes in the cumulus-capped boundary 2008. layer, Q. J. Roy. Meteorol. Soc., 127, 407–421, 2001. Huntingford, C., Jones, R. G., Prudhomme, C., Lamb, R., Gash, Grant, A. L. M. and Brown, A. R.: A similarity hypothesis for J. H. C., and Jones, D. A.: Regional climate-model predictions shallow-cumulus transports, Q. J. Roy. Meteorol. Soc., 125, of extreme rainfall for a changing climate, Q. J. Roy. Meteorol. 1913–1936, 1999. Soc., 129, 1607–1621, 2003. Gregory, D. and Allen, S.: The effect of convective downdraughts Jenkinson, D. S., Andrew, S. P. S., Lynch, J. M., Goss, M. J., and upon NWP and climate simulations, in: Nineth conference on Tinker, P. B.: The turnover of organic carbon and nitrogen in soil, numerical weather prediction, Denver, Colorado, 122–123, 1991. Philos. Trans. Roy. Soc. London, B329, 361–368, 1990. Gregory, D. and Rowntree, P. R.: A massflux convection scheme Johansen, O.: Thermal conductivity of soils, Ph.D. thesis, Univer- with representation of cloud ensemble characteristics and sta- sity of Trondheim, Norway, 1975. bility dependent closure, Mon. Weather Rev., 118, 1483–1506, Jones, A., Roberts, D. L., and Slingo, A.: A climate model study 1990. of indirect radiative forcing by anthropogenic sulphate aerosols, Gregory, D., Kershaw, R., and Inness, P. M.: Parametrization of Nature, 370, 450–453, 1994. momentum transport by convection II: Tests in single-column Jones, A., Roberts, D. L., Woodage, M. J., and Johnson, C. E.: In- and general circulation models, Q. J. Roy. Meteorol. Soc., 123, direct sulphate aerosol forcing in a climate model with an inter- 1153–1183, 1997. active sulphur cycle, J. Geophys. Res., 106, 20293–20310, 2001. Gregory, D., Shutts, G. J., and Mitchell, J. R.: A new gravity-wave- Kurucz, R. L. and Bell, B.: Atomic Line Data, CD-ROM 23, drag scheme incorporating anisotropic orography and low-level Harvard Smithsonian Center for Astrophysics, Cambridge, MA,

www.geosci-model-dev.net/4/919/2011/ Geosci. Model Dev., 4, 919–941, 2011 940 D. N. Walters et al.: MetUM GA3.0/3.1 and JULES GL3.0/3.1 configurations

1995. Minobe, S., Kuwano-Yoshida, A., Komori, N., Xie, S.-P., and Lean, H. W.: Simulation of orographic rainfall in South Wales with Small, R. J.: Influence of the Gulf Stream on the troposphere, UM Version 5.2, Tech. Rep. 70, JCMM, Met Office/University Nature, 452, 206–209, 2008. of Reading Joint Centre for Mesoscale Meteorology, Reading, Mlawer, E. J., Clough, S. A., Brown, P. D., and Tobin, D. C.: Recent 2002. developments in the water vapor continuum, in: Nineth ARM Lean, H. W., Clark, P. A., Dixon, M., Roberts, N. M., Fitch, A., Science Team meeting, San Antonio, Texas, 22-26th March Forbes, R., and Halliwell, C.: Characteristics of high-resolution 1999, edited by: Burleigh, N. and Carrothers, D., 1–6, 1999. versions of the Met Office Unified Model for forecasting convec- Morcrette, C. J. and Petch, J. C.: Analysis of prognostic cloud tion over the United Kingdom, Mon. Weather Rev., 136, 3408– scheme increments in a climate model, Q. J. Roy. Meteorol. Soc., 3424, 2008. 136, 2061–2073, 2010. Lean, J.: Evolution of the Sun’s spectral irradiance since the Maun- Nachtergaele, F., van Velthuizen, H., Verelst, L., Batjes, N., Dijk- der Minimum, Geophys. Res. Lett., 27, 2425–2428, 2000. shoorn, K., van Engelen, V., Fischer, G., Jones, A., Montanarella, Legates, D. R. and Willmott, C. J.: Mean seasonal and spatial vari- L., Petri, M., Prieler, S., Teixeira, E., Wiberg, D., and Shi, X.: ability in global surface air temperature, Theor. Appl. Climatol., Harmonized World Soil Database (version 1.0), FAO, Rome, 41, 11–21, 1990. Italy and IIASA, Laxenburg, Austria, 2008. Lock, A. P.: The numerical representation of entrainment in Nicholls, S.: The dynamics of stratocumulus: Aircraft observations parametrizations of boundary layer turbulent mixing, Mon. and comparisons with a mixed layer model, Q. J. Roy. Meteorol. Weather Rev., 129, 1148–1163, 2001. Soc., 110, 783–820, 1984. Lock, A. P., Brown, A. R., Bush, M. R., Martin, G. M., and Smith, Oki, T.: Validating the runoff from LSP-SVAT models using a R. N. B.: A new boundary layer mixing scheme. Part I: Scheme global river routing network by one degree mesh, in: AMS 13th description and single-column model tests, Mon. Weather Rev., Conference on Hydrology, Long Beach, California, 2–7th Febru- 128, 3187–3199, 2000. ary 1997, 319–322, 1997. Madec, G.: NEMO ocean engine, Institut Piere-Simon Laplace Oki, T. and Sud, Y. D.: Design of Total Runoff Integrating Pathways (IPSL), France, No. 27, ISSN No. 1288–1619, 2008. (TRIP) – A global river channel network, Earth Interact., 2, 1–36, Manners, J.: Radiative transfer over resolved topographic features 1998. for high resolution weather prediction, Q. J. Roy. Meteorol. Soc., Pincus, R., Barker, H. W., and Morcrette, J. J.: A fast, flex- in press, doi:10.1002/qj.956, 2011. ible, approximate technique for computing radiative transfer Manners, J., Thelen, J.-C., Petch, J., Hill, P., and Edwards, J. M.: in inhomogeneous cloud fields, J. Geophys. Res., 108, 4376, Two fast radiative transfer methods to improve the temporal sam- doi:10.1029/2002JD003322 , 2003. pling of clouds in numerical weather prediction and climate mod- Posselt, R. and Lohmann, U.: Introduction of prognostic rain in els, Q. J. Roy. Meteorol. Soc., 135, 457–468, 2009. ECHAM5: design and single column model simulations., At- Marshall, J. S. and Palmer, W. M.: The distribution of raindrops mos. Chem. Phys., 8, 2949–2963, doi:10.5194/acp-8-2949-2008, with size, J. Meteor., 5, 165–166, 1948. 2008. Martin, G. M., Ringer, M. A., Pope, V. D., Jones, A., Dearden, C., Rais¨ anen,¨ P., Barker, H. W., Khairoutdinov, M. F., Li, J., and and Hinton, T. J.: The physical properties of the atmosphere in Randall, D. A.: Stochastic generation of subgrid-scale cloudy the new Hadley Centre Global Environment Model (HadGEM1). columns for large-scale models, Q. J. Roy. Meteorol. Soc., 130, Part I: Model description and global climatology, J. Climate, 19, 2047–2067, 2004. 1274–1301, 2006. Rawlins, F., Ballard, S. P., Bovis, K. J., Clayton, A. M., Li, D., Martin, G. M., Milton, S. F., Senior, C. A., Brooks, M. E., Ineson, Inverarity, G. W., Lorenc, A. C., and Payne, T. J.: The Met Office S., Reichler, T., and Kim, J.: Analysis and reduction of system- global four-dimensional variational data assimilation scheme, Q. atic errors through a seamless approach to modeling weather and J. Roy. Meteorol. Soc., 133, 347–362, 2007. climate, J. Climate, 23, 5933–5957, 2010. Redelsperger, J.-L., Guichard, F., and Mondon, S.: A parametriza- McCabe, A. and Brown, A. R.: The role of surface heterogeneity tion of mesoscale enhancement of surface fluxes for large-scale in modelling the stable boundary layer, Bound.-Lay. Meteorol., models, J. Climate, 13, 402–421, 2000. 122, 517–534, 2007. Rothman, L. S., Barbe, A., Benner, D. C., Brown, L. R., Camy- Mercado, L. M., Huntingford, C., Gash, J. H. C., Cox, P. M., and Peyret, C., Carleer, M. R., Chance, K., Clerbaux, C., Dana, V., Jogireddy, V.: Improving the representation of radiative intercep- Devi, V. M., Fayt, A., Flaud, J.-M., Gamache, R. R., Goldman, tion and photosynthesis for climate model applications, Tellus, A., Jacquemart, D., Jucks, K. W., Lafferty, W. J., Mandin, J.- B59, 553–565, 2007. Y., Massie, S. T., Nemtchinov, V., Newnham, D. A., Perrin, A., Milton, S., Brooks, M., Lock, A., Whelan, E., and Wilson, D.: Rinsland, C. P., Schroeder, J., Smith, K. M., Smith, M. A. H., HadGEM1 physics for the global NWP model (Cycle G34) im- Tang, K., Toth, R. A., Auwera, J. V., Varanasi, P., and Yoshino, provements to boundary layer, large scale precipitation, convec- K.: The HITRAN molecular spectroscopic database: edition of tion and Saharan albedo, Tech. Rep. 458, Forecasting R&D, Met 2000 including updates through 2001, J. Quant. Spectrosc. Ra., Office, Exeter, UK, 2005. 82, 5–44, 2003. Milton, S. F., Greed, G., Brooks, M. E., Haywood, J., Johnson, Sachidananda, M. and Zrnic,´ D. S.: Differential propagation phase B., Allan, R. P., Slingo, A., and Grey, W. M. F.: Modeled and shift and rainfall rate estimation, Radio Sci., 21, 235–247, 1986. observed atmospheric radiation balance during the West African Scaife, A. A., Butchart, N., Warner, C. D., and Swinbank, R.: Im- dry season: Role of mineral dust, biomass burning aerosol, and pact of a spectral gravity wave parametrization on the strato- surface albedo, J. Geophys. Res., 113, D00C02, 2008. sphere in the Met Office Unified Model, J. Atmos. Sci., 59,

Geosci. Model Dev., 4, 919–941, 2011 www.geosci-model-dev.net/4/919/2011/ D. N. Walters et al.: MetUM GA3.0/3.1 and JULES GL3.0/3.1 configurations 941

1473–1489, 2002. Valcke, S.: OASIS3 user guide (prism 2-5), Tech. Rep. 3, Pro- Scaife, A. A., Copsey, D., Gordon, C., Harris, C., Hinton, T., Kee- gramme for integrated earth system modelling (PRISM) support ley, S., O’Neill, A., Roberts, M., and Williams, K.: Improved initiative, last access: 12 April 2011, 2006. Atlantic blocking in a climate model, Geophys. Res. Lett., sub- van Genuchten, M. T.: A closed-form equation for predicting the mitted, 2011. hydraulic conductivity of unsaturated soils, Soil Sci. Soc. Am. Senior, C. A., Arribas, A., Brown, A. R., Cullen, M. J. P., Johns, J., 44, 892–898, 1980. T. C., Martin, G. M., Milton, S. F., Smith, D. M., Webster, S., Viterbo, P., Beljaars, A. C. M., Mahouf, J.-F., and Teixeira, J.: The and Williams, K. D.: Synergies between numerical weather pre- representation of soil moisture freezing and its impact on the sta- diction and general circulation climate models, in: The develop- ble boundary layer, Q. J. Roy. Meteorol. Soc., 125, 2401–2426, ment of atmospheric general circulation models, edited by: Don- 1999. ner, L., Schubert, W., and Somerville, R., Cambridge University Warner, C. D. and McIntyre, M. E.: An ultrasimple spectral Press, Cambridge, UK, 2010. parametrization for nonorographic gravity waves, J. Atmos. Sci., Shaffrey, L. C., Stevens, I., Norton, W. A., Roberts, M. J., Vidale, 58, 1837–1857, 2001. P.-L., Harle, J. D., Jrrar, A., Stevens, D. P., Woodage, M. J., Webster, S., Brown, A. R., Cameron, D. R., and Jones, C. P.: Im- Demory, M.-E., Donners, J., Clarke, D. B., Clayton, A., Cole, provements to the representation of orography in the Met Office J. W., Wilson, S. S., Connolley, W. M., Davies, T. M., Iwi, A. M., Unified Model, Q. J. Roy. Meteorol. Soc., 129, 1989–2010, 2003. Johns, T. C., King, J. C., New, A. L., Slingo, J. M., Slingo, A., Wells, H., Webster, S., and Brown, A.: The effect of rotation on Steenman-Clark, L., and Martin, G. M.: U.K. HiGEM: the new the pressure drag force produced by flow around long mountain U.K. High-resolution Global Environment Model – Model de- ridges, Q. J. Roy. Meteorol. Soc., 131, 1321–1338, 2005. scription and basic evaluation, J. Climate, 22, 1861–1896, 2009. Wentz, F. J. and Spencer, R. W.: SSM/I rain retrievals within a uni- Smith, D. M., Cusack, S., Colman, A. W., Folland, C. K., Harris, fied all-weather ocean algorithm, J. Atmos. Sci., 55, 1613–1627, G. R., and Murphy, J. M.: Improved surface temperature predic- note: SSM/I data are produced by Remote Sensing Systems and tion for the coming decade from a global climate model, Science, sponsored by the NASA Earth Science MEaSUREs DISCOVER 317, 769–799, 2007. Project. Data are available at www.remss.com, last access: 12 Smith, R. N. B.: A scheme for predicting layer cloud and their water April 2011, 1998. content in a general circulation model, Q. J. Roy. Meteorol. Soc., Wielicki, B. A., Barkstrom, B. R., Harrison, E. F., Lee, R. B., Smith, 116, 435–460, 1990. G. L., and Cooper, J. E.: Clouds and the Earth’s Radiant En- Staniforth, A., Wood, N., and Cotˆ e,´ J.: A simple comparison of ergy System (CERES): An Earth observing system experiment, four physics-dynamics coupling schemes, Mon. Weather Rev., B. Am. Meteorol. Soc., 77, 853–868, 1996. 130, 3129–3135, 2002. Wilson, D. R. and Ballard, S. P.: A microphysically based precipi- Stratton, R. A., Stirling, A., and Derbyshire, S.: Changes tation scheme for the UK Meteorological Office Unified Model, and developments to Convective Momentum Transport (CMT) Q. J. Roy. Meteorol. Soc., 125, 1607–1636, 1999. parametrization based on analysis of CRM and SCM, Tech. Rep. Wilson, D. R., Bushell, A. C., Kerr-Munslow, A. M., Price, J. D., 530, Forecsating R&D, Met Office, Exeter, UK, 2009. and Morcrette, C. J.: PC2: A prognostic cloud fraction and con- Taylor, K. E., Stouffer, R. J., and Meehl, G. A.: A Summary of the densation scheme. I: Scheme description., Q. J. Roy. Meteorol. CMIP5 Experiment Design, http://cmip-pcmdi.llnl.gov/cmip5/ Soc., 134, 2093–2107, 2008a. docs/Taylor CMIP5 design.pdf, last access: 12 April 2011, Wilson, D. R., Bushell, A. C., Kerr-Munslow, A. M., Price, J. D., 2009. Morcrette, C. J., and Bodas-Salcedo, A.: PC2: A prognostic Tripoli, G. J. and Cotton, W. R.: A numerical investigation of sev- cloud fraction and condensation scheme. II: Climate model sim- eral factors contributing to the observed variable intensity of deep ulations., Q. J. Roy. Meteorol. Soc., 134, 2109–2125, 2008b. convection over south Florida, J. Appl. Meteorol., 19, 1037– Wood, N., Diamantakis, M., and Staniforth, A.: A monotonically- 1063, 1980. damping second-order-accurate unconditionally-stable numeri- Untch, A. and Simmons, A. J.: Increased stratospheric resolu- cal scheme for diffusion, Q. J. Roy. Meteorol. Soc., 133, 1559– tion in the ECMWF forecasting system, ECMWF Newsletter 82, 1573, 2007. ECMWF, Reading, UK, 1999. Xie, P. and Arkin, P. A.: Global precipitation: A 17-year monthly Uppala, S. M., Kallberg,˚ P. W., Simmons, A. J., Andrae, U., V. Da analysis based on gauge observations, satellite estimates, and nu- Costa Bechtold, Fiorino, M., Gibson, J. K., Haseler, J., Hernan- merical model outputs, B. Am. Meteorol. Soc., 78, 2539–2558, dez, A., Kelly, G. A., Li, X., Onogi, K., Saarinen, S., Sokka, 1997. N., Allan, R. P., Andersson, E., Arpe, K., Balmaseda, M. A., Zhong, W. and Haigh, J. D.: An efficient and accurate correlated-k Beljaars, A. C. M., van de Berg, L., Bidlot, J., Bormann, N., parameterization of infrared radiative transfer for troposphere- Caires, S., Chevallier, F., Dethof, A., Dragosavac, M., Fisher, stratosphere-mesosphere GCMs, Atmos. Sci. Lett., 1, 125–135, M., Fuentes, M., Hagemann, S., Holm,´ E., Hoskins, B. J., Isak- 2000. sen, L., Janssen, P. A. E. M., Jenne, R., Mcnally, A. P., Mahfouf, Zhong, W., Osprey, S. M., Gray, L. J., and Haigh, J. D.: In- J.-F., Morcrette, J.-J., Rayner, N. A., Saunders, R. W., Simon, P., fluence of the prescribed solar spectrum on calculations of Sterl, A., Trenberth, K. E., Untch, A., Vasiljevic, D., Viterbo, P., atmospheric temperature, Geophys. Res. Lett., 35, L22813, and Woollen, J.: The ERA-40 re-analysis, Q. J. Roy. Meteorol. doi:10.1029/2008GL035993, 2008. Soc., 131, 2961–3012, 2005.

www.geosci-model-dev.net/4/919/2011/ Geosci. Model Dev., 4, 919–941, 2011